Questions to ask EMC regarding their new VNX systems…

It’s that time of the year again. The usual websites are busy with news of the upcoming EMC midrange refresh called VNX. And records being broken.

(NEWSFLASH: Watching the webcast now, the record they kept saying they would break ended up being some guy jumping over a bunch of EMC arrays with a motorcycle – and here I was hoping to see some kind of performance record…)

I’m not usually one to rain on anyone’s parade, but I keep seeing the “unified” word a lot, but based on what I’m seeing, it’s all more of the same, albeit with newer CPUs, a different faceplate, and (join the club) SAS. I’m sure the new systems will be faster courtesy of faster CPUs, more RAM and SAS. But are they offering something materially closer to a unified architecture?

Note that I’m not attacking anything in the EMC announcement, merely the continued “unified” claim. I’m sure the new Data Domain, Isilon and Vmax systems are great.

So here are some questions to ask EMC regarding VNX – I’ll keep this as a list instead of a more verbose entry to keep things easy for the ADD-afflicted and allow easier copy-paste into emails 🙂

  1. Let’s say I have a 100TB VNX system. Let’s say I allocate all 100TB to NAS. Then let’s say that all the 100TB is really chewed up in the beginning but after a year my real data requirements are more like 70TB. Can I take that 30TB I’m not using any more and instantly use it for FC? Since it’s “unified” and all? Without breaking best practices for LUN allocation to Celerra? Or is it forever tied to the NAS part and I have to buy all new storage if I don’t want to destroy what’s there and start from scratch?
  2. Is the VNX (or even the NS before it) 3rd-party verified as an over 5-nines system? (I believe the CX is but is the CX/NS combo?)
  3. How is the architecture of these boxes any different than before? It looks like you still have 2 CX SPs, then some NAS gateways. Seems like very much the same overall architecture and there’s (still) nothing unified about it. I call for some truth in advertising! Only the little VNXe seems materially different (not in the software but in the amount of blades it takes to run it all).
  4. Are the new systems licenced by capacity?
  5. Can the new systems use more than the 2TB of FAST Cache?
  6. On the subject of cache, what is the best practice regarding the minimum number of SSDs to use for cache? Is it 8? How many shelves/buses should they be distributed on?
  7. What is the best practice regarding cache oversubscription and how is this sized?
  8. Since the FAST Cache can also cache writes, what are the ramifications if the cache fails? How many customers have had this happen? After all, we are talking about SSDs, and even mirrored SSDs are much less reliable than mirrored RAM.
  9. What’s the granularity for using RecoverPoint to replicate the NAS piece? Seems like it needs to replicate everything NAS as one chunk as a large consistency group, with Celerra Replicator needed for more granular replication.
  10. What’s the granularity for recovering NAS with RecoverPoint? Seems like you can’t do things by file or by volume even. The entire data mover may need to be recovered in one go, regardless of the volumes within.
  11. When using RecoverPoint, does one need to not use storage pools for certain operations? And what does that mean regarding the complexity of implementation?
  12. Speaking of storage pools, when are they recommended, when not, and why? And what does that mean about the complexity of administration?
  13. What functionality does one lose if one does not use pools?
  14. Can one prioritize FAST Cache in pool LUNs or is cache simply on or off for the entire pool?
  15. Can I do a data-in-place upgrade from CX3 or CX4 or is this a forklift upgrade?
  16. Why is FASTv2 not recommended for Exchange 2010 and various other DBs?
  17. If Autotiering is not really applicable to many workloads, what is it really good for?
  18. What is the percentage of flash needed to properly do autotiering on VNX? (it’s only 3% on VMAX since it uses a 7MB page, but VNX uses a 1GB page, which is far more inefficient). Why is FAST still at the grossly inefficient 1GB chunk?
  19. Can FAST on the VNX exclude certain time periods that can confuse the algorithms, like when backups occur?
  20. Is file-level FAST still a separate system?
  21. Why does the low-end VNXe not offer FC?
  22. Can I upgrade from VNXe to VNX?
  23. Does the VNXe offer FAST?
  24. Can a 1GB chunk span RAID groups or is performance limited to 1 RAID group’s worth of drives?
  25. Why are functions like block, NAS and replication still in separate hardware and software?
  26. Why are there still 2 kinds of snapshotting systems?
  27. Are the block snaps finally without a huge write performance impact? How about the NAS snaps?
  28. Are the snaps finally able to be retained for years if needed?
  29. Why are there 4 kinds of replication? (Mirrorview, Celerra Replicator, Recoverpoint, SAN copy)
  30. Why are there still all these OSes to patch? (Win XP in the SPs, Linux on the Control Station and RecoverPoint, DART on the NAS blades, maybe more if they can run Rainfinity and Atmos on the blades as well)
  31. Why still no dedupe for FC and iSCSI?
  32. Why no dedupe for memory and cache?
  33. Why not sub-file dedupe?
  34. Why is Celerra still limited to 256TB per data mover?
  35. Is Celerra still limited to 16TB per volume? Or is yet another, completely separate system (Isilon) needed to do that?
  36. Is Celerra still limited to not being able to share a volume between data movers? Or is, again, Isilon needed to do that?
  37. Can Celerra non-disruptively move CIFS and NFS volumes between data movers?
  38. Why can there not be a single FCoE link to transfer all the protocols if the boxes are “unified”?
  39. Have the thin provisioning performance overheads been fixed?
  40. Have the pool performance bottlenecks been fixed? Or is it still recommended to use normal RAID LUNs for highest performance?
  41. Can one actually stripe/restripe within a FLARE pool now? When adding storage? With thin provisioning?
  42. What is the best practice for expanding, say, a 50 drive pool? How many drives do I have to expand by? Why?
  43. Does one still need to do a migration to use thin provisioning?
  44. Does one need to do yet another migration to “re-thin” a LUN once it gets temporarily chunky?
  45. Have the RAID5 and RAID6 write inefficiencies been fixed? And how?
  46. Will the benchmarks for the new systems use RAID6 or will they, again, show RAID10? After all, most customers don’t deploy RAID10 for everything, and RAID5 is thousands of times less reliable than RAID6. How about some SPC-1 benchmarks?
  47. Why is EMC still not fessing up to using a filesystem for their new pools? Maybe because they keep saying doing so is not a “real” SAN, even in recent communication?
  48. Since EMC is using a filesystem in order to get functionality in the CX SPs like pools, thin provisioning, compression and auto-tiering (and probably dedupe in the future), how are they keeping fragmentation under control? (how the tables have turned!)

What I notice is a lack of thought leadership when it comes to technology innovation – EMC is still playing catch-up with other vendors in many important architectural areas,  and keeps buying companies left and right to plug portfolio holes. All vendors play catch-up to some extent, the trick is finding the one playing catch-up in the fewest areas and leading in the most, with the fewest compromises.

Some areas of NetApp leadership to answer a question in the comments:

  • First Unified architecture (since 2002)
  • First with RAID that has the space efficiency of RAID5, the performance of RAID10 and the reliability of RAID6
  • First with block-level deduplication for all protocols
  • FIrst with zero-impact snapshots
  • First with Megacaches (up to 16TB cache per system possible)
  • First with VMware integration including VM clones
  • First with space- and time-efficient, integrated replication for all protocols
  • First with snapshot-based archive storage (being able to store different versions of your data for years on nearline storage)
  • First with Unified Connect and FCoE – single cable capability for all protocols (FC, iSCSI, NFS, CIFS)

However, EMC is strong when it comes to marketing, messaging and – wait for it – the management part. Since it’s amazingly difficult to integrate all the technologies EMC has acquired over the years (heck, it’s taking NetApp forever to properly integrate Spinnaker and that’s just one other architecture), EMC is focusing instead on the management of the various bits (the current approach being Unisphere, tying together a subset of EMC’s acquisitions).

So, Unified Storage in EMC-speak really means unified management. Which would be fine if they were upfront about it. Somehow, “our new arrays with unified management but not unified architecture” doesn’t quite roll off the tongue as easily as “unified storage”.

Mike Riley eloquently explains whether it’s easier to fix an architecture or fix management here. Ultimately, unified management can’t tackle all the underlying problems and limitations, but it does allow for some very nice demos.

A cool GUI with frankenstorage behind it is like putting lipstick on a pig, or putting a nice shell on top of a car cobbled together from disparate bits. The underlying build is masked superficially, until it’s not… usually, at the worst possible time.

Sure, ultimately, management is what the end user interfaces with. Many people won’t really care about what goes on inside, nor have the time or inclination to learn. I merely invite them to start thinking more about the inner bits, because when things get tricky is also when something like a portal GUI meshing 4-5 different products together also stops working as expected, and that’s also when you start bouncing between 3-4 completely different support teams all trying to figure out which of the underlying products is causing the problem.

Always think in terms of what happens if something goes wrong with a certain subsystem and always assume things will break – only then can you have proper procedures and be prepared for the worst.









And always remember that the more complex a machine, the more difficult it can be to troubleshoot and fix when it does break (and it will break – everything does). There’s no substitute for clean and simple engineering.

Of course, Rube Goldberg-esque machines can be entertaining… if entertainment is what you’re after 🙂



Technorati Tags: , , , , , , , , , , ,


63 Replies to “Questions to ask EMC regarding their new VNX systems…”

  1. Full Disclosure: previously working for a NetApp reseller (NCDA, NCIE), now working for an EMC reseller — doing VMware stuff just the same as I used to. 🙂 I like storage period and am enjoying going really deep on stuff…fun technology regardless of where you sit vendor-wise. So….

    How many management consoles do you need for the underlying unified OS? Systems Manager, CLI (advanced dedup, thin provisioning, lots of “options”, etc.), FilerView (times 2…legacy I realize but still SnapMirror in there), Operations Manager, Protection Manager/Performance Advisor/Provisioning Manager (wrapped into the same GUI but pretty different inside that), SnapManager per SQL/Exchange/Hyper-V server, SMVI/VSC for VMware, SnapDrive, phew. On the other side, you basically have Unisphere, some CLI, and Replication Manager (why in the world doesn’t SnapManager have a unified interface especially for larger multi-server installs?). When installing NetApp, I actually looked over at Unisphere a bit jealously to be honest. What I find ironic is that you have fewer OSes with more management consoles vs. more OSes with fewer management consoles….go figure.

    # of parts under the covers — no disputing that…the big difference I’ve seen personally is that while upgrades can be more complex (depends on scenarios, etc.) the customer ultimately doesn’t care very much as EMC Professional Services will come onsite to do them (included with support). By comparison, upgrades on the NetApp side even if easier are either self-service or a paid PS engagement. For something like a storage array (i.e. super-criticial), a lot of customers just got very nervous about upgrading it themselves.

    Data migration – really just about FC backend shelves to SAS backend shelves — not a lot different I think than no data inplace upgrades when moving from FC shelves (DS14) over to SAS shelves (DS4243, etc.).

    Granularity — can SnapMirror roll back LUNs on an I/O by I/O basis or only to certain checkpoints? Can it roll forward and roll back and roll forward again? Or once you revert to a snapshot have you reverted forevermore? (ah….now we get into questions about volume-level SnapRestore vs. file-level SnapRestore and the resulting snapshot impact)

    Lots more for later….may even have to blog on this myself (just started hitting on some storage array failover details in my last post actually).

  2. Full Disclosure: previously working for a NetApp reseller (NCDA, NCIE), now working for an EMC reseller — doing VMware stuff just the same as I used to. 🙂 I like storage period and am enjoying going really deep on stuff…fun technology regardless of where you sit vendor-wise. So….

    How many management consoles do you need for the underlying unified OS? Systems Manager, CLI (advanced dedup, thin provisioning, lots of “options”, etc.), FilerView (times 2…legacy I realize but still SnapMirror in there), Operations Manager, Protection Manager/Performance Advisor/Provisioning Manager (wrapped into the same GUI but pretty different inside that), SnapManager per SQL/Exchange/Hyper-V server, SMVI/VSC for VMware, SnapDrive, phew. On the other side, you basically have Unisphere, some CLI, and Replication Manager (why in the world doesn’t SnapManager have a unified interface especially for larger multi-server installs?). When installing NetApp, I actually looked over at Unisphere a bit jealously to be honest. What I find ironic is that you have fewer OSes with more management consoles vs. more OSes with fewer management consoles….go figure.

    # of parts under the covers — no disputing that…the big difference I’ve seen personally is that while upgrades can be more complex (depends on scenarios, etc.) the customer ultimately doesn’t care very much as EMC Professional Services will come onsite to do them (included with support). By comparison, upgrades on the NetApp side even if easier are either self-service or a paid PS engagement. For something like a storage array (i.e. super-criticial), a lot of customers just got very nervous about upgrading it themselves.

    Data migration – really just about FC backend shelves to SAS backend shelves — not a lot different I think than no data inplace upgrades when moving from FC shelves (DS14) over to SAS shelves (DS4243, etc.).

    Granularity — can SnapMirror roll back LUNs on an I/O by I/O basis or only to certain checkpoints? Can it roll forward and roll back and roll forward again? Or once you revert to a snapshot have you reverted forevermore? (ah….now we get into questions about volume-level SnapRestore vs. file-level SnapRestore and the resulting snapshot impact)

    Lots more for later….may even have to blog on this myself (just started hitting on some storage array failover details in my last post actually).

  3. Hi Andrew,

    Since you have some NetApp experience you must also know that the stuff in Operations Manager/Protection Manager/Provisioning manager is not to be found within Unisphere, if indeed anywhere else in the EMC universe. More akin to Ionix probably. And it comes with all new systems (how much is Ionix again?)

    The complexity with VNX (AKA Celerra) is not just troublesome for upgrades, which, though more time-consuming, should be a fairly routine procedure.

    The point I tried to make in my post is that having all that extra stuff behind the curtain can get problematic. I did the opposite switch you did – from selling EMC to selling NetApp. I felt Celerra was a nightmare when things didn’t go right, and the separation between the disparate bits making up the thing never feels clearer than when you’re having problems and the bits aren’t talking to each other properly. It’s a house of cards.

    Oh, and new NetApp systems can absolutely use the old FC shelves (with a data in place upgrade) AND the new SAS ones. VNX can’t use existing FC shelves, no matter how large the customer investment.

    Providing a single pane of glass management is nice (as long as the functionality doesn’t suffer) and is more an exercise in HOW to present the information than anything else. It’s doable. EMC has taken the first step with Unisphere (which, as you well know, is missing a ton of EMC products at the moment).

    Fixing the back-end gubbins, on the other hand, is very, very tricky indeed.

    Recoverpoint is indeed nice and granular – for LUNs. But it does have a lot of extra I/O and space overhead to account for, especially for CDP. See


    1. Sorry for the extremely long delay here…trip out of country for vacation + getting sick + workload meant I haven’t been back here for a while…crazy how many comments there are now.

      Just so it looks I didn’t run out on the discussion… 😉

      -OM/PM/etc. — yes, I agree some very cool things there (I spent a lot of time mucking around OM and kept finding new things in it). What was interesting was how few customers in the channel space actually dug into OM/PM/etc (even after we did the setup and/or walkthroughts). I’ll agree it goes far further than Unisphere does (we’re into Ionix territory I agree) but there were a lot of times I wanted a basic form of what OM/PM provide right inside FilerView/System Manager (which Unisphere does provide).
      -Behind the curtain — this is something I’ll still somewhat open-minded about to be honest. At the end of the day, I try to focus on solving the customer needs (and am fairly comfortable I can do so 90%+ of the time regardless of which vendor to be honest). Having said that, we’re talking implementation and/or code problems…which do over time get better (lots of improvements in FLARE/DART over the years…some competitive talking points I used to know aren’t true anymore for instance).
      -True on shelves overall — but also true that can’t retain internal 20×0 disks when upgrading to a 31×0 (why I’d often try to sell empty 20×0 systems…the catch was that internal 20×0 disks were by far the cheapest disk options…that really makes a difference in the channel).
      -Single pane of glass — agreed it may be easier….let’s have NetApp do it too please. 😀 (oh, how I wanted System Manager for a long time….and then wanted it to do so much more too although even as a 1.0 release it was very stable)
      -RecoverPoint — now for NFS too. 😀 And yes, it has more overhead…but does provide the ability to do so much more too. 🙂 If you’re using heavily using snapshots, I’d posit you’re to the point where you have enough feature interaction (snapshots vs. dedup vs. single-file restore vs. clones vs. full-volume rollback removing newer snapshots vs. etc.) that the RP overhead actually isn’t bad.

      Thanks for the discussion…on a flight back from Singapore this weekend I just happened to sit across the Principal Virtualization Consultant for HP in the APJ area…although we do very different storage it was a fun conversation on and off during the flight…would love to have that same kind of wandering conversation with you as well sometime (kind of how Vaughn and Chad can get along overall…they agree on 80% of their tech worldview as Vaughn said to me one time).

      1. Hi Andrew, hope you had a nice vacation and that you’re feeling better.

        Granted regarding upgrading from a 2040 to a 3xxx – you do lose the internal disks ONLY. That’s 12 drives. But with the VNX systems you cannot use ANY of your old stuff. I would have thought EMC would have imitated NetApp and allowed simultaneous SAS and FC shelves but this is a case where they didn’t. Most NetApp and EMC customers have way more than 12 drives…

        And at least you CAN upgrade from a 2040 to a larger NetApp system, you could never upgrade from an AX4/NX4/VNXe to CX/NS/VNX. Yet another silo. I have customers that were on NetApp 2000-series boxes, and once they outgrew them they were able to upgrade to, say, a 3210 w/ Flash Cache for a nice bump in speed and scalability for not much more money.

        And regarding RecoverPoint for NFS – do read the documentation, it still has no concept of NFS and just replicates the CX LUNs given to a Celerra head. This is nothing new, it could always be done, just not advertised… 🙂

        Yet another case of marketing over substance.


  4. Dimitri-

    Thanks for your continued interest in EMC’s products. While I can’t comment on things that haven’t been released (that would be of questionable ethics since I work for EMC), I would caution you around two areas:

    1) When poking at the file system limitation question– Let’s face it, two node architectures generally weren’t designed to handle massively large file systems (think multiple PB+ size or crazy numbers of luns for that matter) that are possible today with aggregation of a massive capacity drives we have in the market (I’m hearing about an 8TB drive on the horizon btw – holy smokes! think of the rebuild times!). Usually customers love their first Filer, but by the time they get to their forth, 100TB file systems and all, their management time often increases linearly. You guys have a host of “Managers” (how many are we up to now :-)??) to help but the underlying issue still remains. The issue I’m talking about is excruciatingly painful in certain verticals. Think of a VERY large music service that many of us use for example and you’ll start to grasp what I mean ☺ when I allude to certain parts of the market needing something entirely different. This is why EMC now owns Isilon (sorry I’m sure that has to sting just a bit). AND.. This is why you guys have Spinnaker right?..uhhem..I mean GX..umhem, I mean “Ontap 8”? Sorry it’s hard to follow sometimes and the Ontap 8 (lowercase c mode?) name has many even more confused these days. Is that naming convention by design by marketing or should I assume ontap 8c mode can seamlessly replicate with another “Ontap 8” non c mode system without issue? Heck, can I upgrade form Ontap 7 to Ontap 8-c mode and keep mind data in place? After all they both share the same name, right? Seems like a reasonable assumption for a customer to make. Now Dimitri, notice I’m by no means saying it’s BAD to have a different tool in the tool chest or even a combo of different technologies blended together to form an integrated solution (think: Bycast object storae in front of ntap filers).

    2) Okay.. I’m no mind reader BUT..It seems like you might be assuming that when EMC competes with any mid tier competitor they limit themselves to a one product solution (let’s not get into semantics around number of heads’s getting tired) to solve all the customer’s needs? Rather than starting with thinking what is possible inside one system, then trying figure out how to force it into a given use case, perhaps it would be better to use the best tool for the job and perhaps even a blended solution with a central management framework? This is what your sort of “parent company” IBM has been doing for many years with decent success. Do they have it wrong somehow? EMC (or IBM or Oracle etc) will NEVER have the mindset of “If this one product can’t do it, you don’t need it”. I hear we (EMC) did a long time ago and it didn’t work out too well :-). And let’s be honest, I don’t think NetApp does either anymore since you have 3 separate and very distinct storage operating systems (marketing and naming conventions aside). On the multiple “moving parts” question being inherently bad for upgrades – (which you will see shortly isn’t an issue at all in vnx or arguably ever was in NS) – does this mean that Bycast in front of a set of filers, with various file system structures layered on top of one another is an inherently less reliable solution for someone who wants object storage than an integrated Atmos with built in storage? If so, then I’d expect you to always position Atmos in those use cases exclusively :-).

    Oh yeah..before I forget.. if you want to learn more, please reach out to me separately. I have a great way for you to “get more insider/pre-release info”, it just might mean not working at NetApp anymore though 😉


  5. Well said Jonas…..

    Dmitri – I always enjoy good journalism and freedom of speech so long as there is 100% accuracy in the content. I too can’t comment on this new platform but might have plenty to say tomorrow 🙂 Much like Andrew, I made the journey from an IBM Business Partner reselling NetApp to evangelizing EMC’s capabilities. I can tell you during my time at this IBM Business Partner and prior, I had and still have a lot of respect for our Sunnyvale friends.

    Before we dive, can you define in NetApp’s term what innovation is and why you might consider yourself (or NetApp) more of an innovator than others in this space? I might have a different view and would love to get your opinion since you did mention this above. It is no secret that 2010 was the year of “Acquisitions” within our industry. Whether you want to admit it or not, innovation can be spawned through acquisition and often does. We are certainly not shy on internal innovation efforts and have plenty of areas to collaborate on as you are probably well aware of.

    Would you consider the following within your “Thought Leadership” category?


    Yes, we have made plenty of acquisitions just like NetApp and curious why you think NetApp is at the top of the innovation ladder and why EMC is any less of an innovator? I for one think the Spinnaker acquisition was a good acquisition but didn’t have any idea that it would take NetApp over 6+ years to bring a fully functional solution to market (still not fully functional). On the contrary, EMC has a long and successful acquisition history resulting in product releases shortly after acquisition (think DCA – Data Computing Appliance – 70 days after acquisition).

    I also think of innovation around market trends. Much like Apple who I personally believe is at the top of the list. They set trends, create demand, and have a phenomenal marketing strategy. If you didn’t see the market shift then perhaps the crazy M&A activity might be a stark reminder. Some notables, EMC buys Isilon, Dell buys Compellent, HP buys 3Par. All of these acquisitions were made (with the exception of the Isilon purchase) in response to VCE. Oh, did I mention NetApp cobbling together “flexpod” in an attempt to join the club? The idea of providing a fully integrated stack from management, provisioning, and support sounds extremely appealing and is what customers have been asking for. The big question is, can all of these components be integrated, managed, and provisioned through a simple and easy interface while still holding true to the single pane of glass. NOT a bunch of element managers but a single pane of glass. I firmly believe NetApp will get to a Unified Management framework, let’s just hope it doesn’t take as long as the Spinnaker development.

    Tomorrow is a BIG day, make sure you get some rest as there is lots to catch up on. I expect through the coming weeks you will have some answers and most likely have to make some changes to your post.

    Looking forward to future conversations….

  6. @Aaron: Indeed I made some changes to my post, I’d forgotten some questions. I watched the webcast, I read all the materials, and the questions remain, with additions.

    I did add another piece regarding NetApp thought leadership – look for the bulleted list. That plus the entire way NetApp systems are sold, including software bundles, is what EMC is aspiring to reach (the new EMC software bundles are eerily similar to how NetApp has been selling systems for a while now).

    Regarding acquisitions:

    NetApp has made a few (nowhere near as many as EMC) and, with the exception of management pieces like Onaro and Akorri, the other tech is stuff we always try to put into our unified codebase – doing otherwise flies in the face of our philisophy.

    Look at Spinnaker (which became GX). Instead of leaving it as a separate product (like EMC does) NetApp has been hard working to integrate ONTAP with GX.

    The codebases are totally different – imagine trying to integrate Enginuity with Celerra. How long would that take? Or Celerra with Isilon.

    NetApp takes the infinitely more difficult path of true integration, not just integration on Powerpoint slides.

    Spinnaker is and has been “fully functional” since it was acquired BTW.

    Eagerly awaiting answers…



  7. @Jonas:

    I understand you left at the beginning of 2007 from NetApp. Almost 4 years ago. I guess then you haven’t played much with, say, Provisioning/Protection Manager.

    The Isilon acquisition doesn’t sting at all – NetApp could have acquired them with spare change years ago, same goes for 3Par etc. We didn’t see a need.

    What I know is that now EMC has yet another totally incompatible system to offer. Not that it doesn’t have its uses, hopefully it’ll be worth the money you spent.

    Unisphere is a small, limited single pane of glass. It doesn’t have the richness of other interfaces, and in many ways it’s a portal interface. It’s not as if Unisphere contains Ionix and all other management tools EMC has. It contains a SUBSET.

    And the “best tool for the job” argument is only valid if, indeed, you have all the best tools to start with.

    1. Email/File Xtender – are you kidding me? Hopefully the new version is working out.
    2. Replication manager – I had so many returned it wasn’t even funny. There were also 2-3 variants (RM-SE, RM etc).
    3. Networker – nobody will even put up a fight on this one. Often given away to cement a deal.
    4. Alphastor?
    5. Autostart?
    6. Invista? (that one was almost comical)
    7. Rainfinity – customers often have performance problems with it, and its largest customers are looking for a way out. F5 ARX is better (I know of a large trading firm that returned Rainfinity and got F5 instead).

    I could go on.

    Just because you bought them doesn’t make them “best of breed”. It makes them something you could afford at the time and that maybe had a future.

    I’ll close with a final thought and then wait for some answers to my questions instead of even more questions:

    I’ve never lost a proof-of-concept against Celerra, with Unisphere or not. The only time I’ve lost (once) was when the Celerra was free. Yes, free. And many other times it was free or close to it and the customer still didn’t buy it.

    People do care about technology. Marketing will get you very far indeed, but at some point a dude with a motorcycle jumping over a bunch of EMC racks is not a compelling record to break.

    Really I was hoping for some Isilon SPEC record or somesuch, maybe finally showing RAID6 working decently.

    Next time?


  8. @Gene: I wish, 90 minutes of my life I’ll never get back! Some Knievel-type dude with an old Harley indeed set a record for “guy that has jumped over most EMC racks”. Alrighty then…

  9. You’re still singing the same tune my friend, it’s just from the wrong sheet of music and doesn’t really form a cohesive argument that works anymore. At least not with the customers I talk to (about ~1000 in 4 years). EMC had an awesome launch..and we had fun at the same time. Those two things aren’t mutual exclusive btw :-). Even if it’s “the biggest product launch” in your company’s history you should have some fun too, no? Otherwise you end up just boring people. Of course, I don’t expect you to concede anything to me:-) as you clearly have the same “fire in the belly” that me, Aaron, Chad and many others @emc and probably ntap have when it comes to competing. It makes for great theater for the market and our long as we keep it “above board”.

    You might want to rethink the comments on Isilon (at least privately). Have a look at the architecture again and compare it to ontap 8-c mode.. Step back from being a ntap super blogger for a minute and look at it as the engineer I know you probably really are. Notice the true cache coherency + a single file system striped wide uniformly across every node, cheap and deep configs combined with ultra fast configs working together in a gigantic grid, insanely simple, FAST like characteristics, massive scale etc.. In terms of what you choose to compete with since you have 3 very separate storage OS products to choose from (or a combo of all three), notice how many answers turn up a host of brick walls that are very tough to circumvent in terms of customer trade-offs. I guess we’ll see what happens in the next 18 months in the market and reconnect on the topic in a bit (who knows?, Dimitri, perhaps I have it wrong. We’ll just have to see how it shakes out :-))

    In terms of the list of emc acquisitions you put forth – I will concede that RM is somewhat still misunderstood and was actually kind of expensive before the new bundles we announced today. I’m a HUGE proponent of the product personally since my background is deeply rooted in the app & db space. I personally have always included it in all my solutions where the customer has any workflow, applications, dbs, development etc…which is pretty much all of them. In the field, I have caught ntap claiming uniqueness with the SnapMgr * (Kroll’s SMBR is my personal fav) tools and if that happens without EMC designing a solution that includes proper app integration, then perhaps you guys deserve the business because our sales team failed the customer (imho). (I know there are a few cool things the SMO product does with ASM btw..kinda unique but that’s not what I’m talking about it here). I’m working feverishly over here to make sure that doesn’t happen though so expect to see it in every deal it where it makes sense. There are some huge advantages to RM over 7 very different snapmgr products but that’s okay ..let’s save that one for another day.

    And by the way, what makes you (and others who’ve used similar rhetoric) think the last time I touched nearly ALL of the ntap products was 4 years ago? Do you possess clairvoyant powers I wasn’t aware of or is that just an unfounded conjecture? My guess is that it’s the latter. You seem to be playing with Unisphere somehow..hmmm.. I won’t try to discredit the hands on experience you have with EMC (although antiquated) just because you blog about it and I may not *like* what you have to say.

    The other companies you mentioned emc bought, may have not been as widely successful as DG, DD, Kaysha, VMware, Avamar, Isilon (will be), etc BUT, I’m fairly confident when I say that the acquisitions you’re pinpointing, only represent a minuscule fraction of the total # of acquisitions EMC has done over the years. BUT what happens when the majority of your acquisitions are abysmal failures with upward of $1B down the drain with nothing to show for it? Heck, you could have had a pretty kick-ass CDP solution by now or even a VTL that was say, #2 or #3 to DD by now with the right level of focus after the initial acquisitions that were made. That suggests to me that there’s ample room for improvement but I only worked at NetApp 4 years ago, right? What do I know 🙂

    A few comments since you threw a few things out there at the end:
    6 POCs for me against you all, (with the inclusion of the 2 management products you mentioned above). We won 5/6 BUT to be fair, (3PAR beat us both on one). It wasn’t limited to Celerra NAS only btw. I got to use a few more assets to solve the challenges put forth by the customer at the outset. Block was also part of the was NAS (in that order). You guys have a great traditional 2 headed NAS. No doubt about it.

    The “giving away” thing you mention happens, and is not unique to EMC btw…I’ve seen it from ntap too on several occasions right here in their own backyard. In fact, while my sample size isn’t scientific it appears to be the ntap M.O. when the deal starts to swing toward EMC. I’ts okay..I get it, sales people get fired up..nobody likes to lose.

  10. Nice post Mr K! I guess EMC has lifted the embargo on commenting on our blogs!!

    The most interesting thing I have observed today is the ‘meh’ reaction from so many in the press and analyst communities. I’d expect any competitor to have such a reaction. Perhaps that is why, as one journalist (Chris Melor??) reports, EMC did not pre-brief the press.

    A knock made by many analysts over the last few years about them is that they are all style and little substance …

    Doing the interesting stuff like unified dedup and dedup cache requires a very mature product. It is not simple, and to do it EMC will have to think differently, but I think that will be the challenge of Sisyphus to them.

    It is easier to market “Just Like NetApp!” …

  11. Hey Jonas,

    Absolutely you should have fun but I was still expecting a performance-related record given that it’s a storage launch event 🙂

    The story behind Isilon and all grid-like products is very nice. The problem is, how do you maintain storage efficiency, advanced features, large addressable space, better than R5 protection, global cache coherency, low latency, high IOPS, high node count and high throughput ALL AT THE SAME TIME?

    Looking at it like the engineer that indeed I am (not part of a competitive team or anything, I do this in my own free time), most scale-out products end up sacrificing something in order to achieve scale-out. That includes ONTAP c-mode.

    I’m sure the Isilon boxes will improve long-term but any Isilon benchmarks I’ve seen show very good sequential I/O but relatively poor random I/O characteristics. Which is fine, it’s what the boxes were designed to do (large datasets, large files). IMO that’s not the hallmark of a good general-purpose system, grid or not. Yes it’s “big data” but for very specific use cases.

    Would be interesting to see how an Isilon grid would do at a random VMware or DB benchmark against NetApp FAS (any protocol), VMAX, VNX FC and VNX NFS.

    But, hey, we’ll see, as you say. This is why I love this industry, always something cool happening, always someone leapfrogging someone else. Who knows, maybe Dell will surprise everyone and crush the scale-out story with Exanet! (whatever happened to that system? Maybe they’re prepping something big)

    Anyway, the post wasn’t about Isilon but about VNX.

    I do indeed possess a modicum of clairvoyance, but I merely based the “out of touch” observation on the fact that you keep mentioning the snapmanagers and not Protection and Provisioning manager and Performance Advisor (tabs under the NetApp Management Console), that come with every NetApp system. Which already goes a very long way into tying things together under a single pane of glass and do complex things no EMC software does yet to my knowledge.

    Plus the NetApp Snapmanagers do some highly advanced things not just with Oracle but also with SQL and Exchange, and let’s not forget VMware. Do watch a recent demo. And I’ll watch a recent RM one and let’s compare notes…

    Oh, and how many RM copies can be taken, and at what space efficiency and what performance overhead? (and if you assume Recoverpoint is used, then that one has its own set of overheads, cool as it is).

    Also, if I’m not mistaken, if someone needs to use a Symm, then there’s no Unisphere and they’re back to managing all the bits separately – right?

    Like with Isilon.

    Like with Data Domain.

    Like with Avamar.


    So it depends on which box you have. Wisely, you picked the most popular one to take care of first. But the moment a customer steps outside that box and needs an additional solution, things start getting complex once more.

    If my experience with EMC is antiquated, then are you willing to touch any of the questions? At least half? They should be real easy! I still see no takers… 🙂 Just more questions from you, that I’m still gamely answering 🙂

    You see, the questions are not an attack on the tech, merely are there to show that, beneath the covers and the (very nice) new bezels, most of the same old limitations are still there for VNX.

    But congrats on all the marketing.


  12. NetApp employee here.

    Another question for EMC: Is it possible for EMC to make an announcement and then not get hugely defensive? I mean, it is O.K. to look at the literature and ask questions and look at the market and see if it’s already been done, right? Can we stop with the “Yeah – well my dad can beat up your dad” debate tactics. EMC, you made the announcement and, unfortunately, it was like getting a bag of chips out of a vending machine – the wrapper looked good but all that was inside was some potato dust and hot air. And you know what – that’s O.K. I saw a blog mention Babe Ruth earlier today. I believe the Babe still holds the record for strike outs, too. You stood in the box and it didn’t turn out the way you wanted. There’s probably another hardware refresh 18 months away. Take another shot when you get the chance. I can tell you, we (NetApp) have been at this Unified Storage thing for about 15 years now. It ain’t easy to build and it’s almost as hard to market.

  13. Good bloody questions;)

    >26. Is Celerra still limited to not being able to share a volume between data movers? Or is, again, Isilon needed to do that?
    Not a problem to end user. Data mover exports Celerra File system to NFS/CIFS at different IP addresses. It doesn’t make sense to access same backend volume from different data mover. If end users want to access NFS/CIFS at a new IP address, just add the IP to data mover’s network interface.

    >27. Why can there not be a single FCoE link to transfer all the protocols if the boxes are “unified”?
    Not a problem to end user.
    Multiple connectivity is for end user choice.
    For end user wanting NAS access, IP based NFS/CIFS is what he wants. For end user wanting SAN access to his host, FC is what he wants.

  14. Dimitri, MikeS & MikeR,

    Please check out Chad’s latest blogs around vnx + vnxe first.
    There’s lots of meat there (many answers to your queries above) and I think in general an emc blog is a better place to hold the authoritative information (I don’t have a blog because I’m not sure the world needs yet another storage blog right now:-)). I have about 50 “questions” for you too, but I wouldn’t be silly and try to jam them into a single blog post and demand you respond in serial fashion..just doesn’t seem like a realistic expectation of any competitor.
    Oh and if you’re interested in dedupe or the impact of EMC v netapp in a VMware environment, please check out a recent EMC customer who was given a ntap vseries (somewhat crassly) right on our emc communities site. The customer actually validated the solution in their datacenter:



  15. Hi Jonas,

    I DID read all the EMC posts regarding VNX and VNXe.

    It’s still clear that VNX is exactly the same as before only with differently structured licensing for software, faster CPUs and more RAM, SAS instead of FC and different marketing. The old limits remain (and I don’t seriously expect answers but I do want customers aware of what to ask – and I know it cuts both ways, which is why I do a ton of research before I post anything).

    VNXe seems to be the one that has a different way of running DART and FLARE – however, every single EMC blog is extremely vague about how all this is done (I read all of Steve Todd’s stuff, Chad’s, Mark’s and Chuck’s). Something about CSX and abstraction – however, it’s not clear whether:

    1. VNXe still runs a Windows kernel
    2. VNXe still runs DART and its kernel
    3. VNXe still runs the control station on top of a Linux kernel.

    If you can confirm any of those it would be grand…

    From the writing it seems that EMC is trying to say they re-wrote FLARE so instead of being a set of WDM drivers running on top of Win XP embedded, it’s now a CSX package running on top of Linux, kinda like Java or Flash can run on various architectures as long as the right hooks are there.

    Is this at all right?

    DART and the control station would be easier since they’re Linux-based anyway…

    My question would be, if this architecture is cleaner, better, faster, easier – why is the VNX proper not based on it?

    Regarding the community message – I saw it. Not sure what box they were trying and whether it had the expanded cache but looking at the 1TB or so of disk space, it doesn’t look like it was exactly given enough disks to play with… and that, anyone with a quarter of a brain should be able to figure out, which is why I actually like that stuff out there.

    Plus that entire comment is highly suspect – if NetApp was really running the PoC the guy wouldn’t be talking about Filerview, redundancy issues and 4-nines when in fact it would be System Manager, fully automated redundancy and a certified well-over 5 nines (approaching six actually). Or 81% efficiency (clearly some kind of RAID5) vs 48% with NetApp, which is only achievable if one doesn’t know what they’re doing or, indeed, if they’re actively trying to stack the deck. Read this:

    Kinda like another PoC I’m aware of that had one of our boxes with a single shelf of disk running performance tests against a large DMX. It’s funny that there wasn’t more of a difference in some tests… 🙂

    I did notice they got 90% dedupe with NetApp…


  16. @Dmitri, MikeS, and Mike R,

    sorry for the delayed response but was enjoying a little time off. We are allowed a little downtime from competing right? 🙂

    I know NetApp is a religion but you should open your eyes to other areas of the market. It amazes me to see that most NetApp fanbois truly believe that you cannot have other complimentary products to a core array. You firmly believe a NetApp FAS platform fits “every” use case and can do everything. I only can extrapolate that this religion believes in no other complimentary technology. A perfect example would be the move away from your VTL technology, to OSSV, to nearly dismantling that effort to then buy DataDomain, and then once EMC snags up DD, snaps were back on the table and the message is, that is the right way to address customers backup challenges. Before, you believed that DD was the right approach and today you don’t? Why so much indecision? Oh, and I forgot, Syncsort is the answer for indexing right? I would be more than happy to talk to your customers about that whole relationship.

    I know MikeR jumped in to ask that we/EMC quit with the “my dad is bigger than your dad” statement. Quite humorous considering all of you have been on the attack trying to deflate the EMC announcement. I consider this to be a healthy debate with the typical NetApp supuriority slant. That’s why we take it to the streets and highlight our competitive advantages. Let’s be honest, does Dmitri’s comment about “I have never lost a POC”. Should we say he didn’t mean to come off a little over confident? There are other words for that but I will let the readers sniff that out. What I can say is that I have never lost to NetApp across many regions so I guess we just haven’t competed against each other. I also don’t spend my time focused on SPC or SPECsfs debates like what NetApp touts because that isn’t a decision criteria. From my experience, customers care about a partnership, they care about how technology enables their business (application integration) both now and over the next 36-48, they care about efficiency and performance (real world), and they want to be able to manage this infrastructure through a simple and easy framework.

    @MikeS, you made a reference to a “meh response in the market” but I can’t help but refer you back to several market responses especially Chris Mellors that stated “NetApp’s launch was like spending millions to develop a new C-Class that looks like the old one”. More of the same. I know it is NetApp’s job, or any competitors, to deflate any momentum stemming from a launch but I am further amazed at NetApp dismissing our launch when their launch was less than impressive. Since Dmitri does such a great job highlighting our launch, let’s use a litte Tivo for storage(RecoverPoint) and rewind (can Netapp do that? Nope didn’t think so). Here is the biggest launch in NetApp’s history:

    Wait for it…. It is impressive… meh..

    * Refresh of the FAS3100 and FAS6000 Platform (pretty standard)
    * SSD support (wow, why did you do that? Is Flashcache the messiah for NetApp? I mean 8TB of MegaCache according to Dmitri. Known Fact –> 2 years behind EMC and the rest of the competitors, why? Oh and I forgot to ask, can I buy just one or two SSD drives? Hmm, I bet the answer is a whole shelf (24 drives). Wow, that isn’t very flexible and quite expensive. Can you intermix the drives in the SAS stack? Didn’t think so… Can you mix different drive types in the same aggregate? Hmm.. bummer. Still a very flexible solution though huh? To better leverage all your drive types can you move data between tiers in a FAST like manner? Wow, you can’t do that either. Weird, still very flexible right?)
    * Flexpod (Really? Are you really going to say this in an integrated stack? Single management? Sure if you want UCSM, System Manager, Operations Manager, Cisco Network Element Manager, Snap Manager, Provision Manager,etc… Pre-packaged? – Hardly!! Single Support, NOT!! only through alliances and considerably different from the VCE model, ALL in response to the tremendous success that VCE is having. A funny blog mentioned that NetApp was selling every solution as a Flexpod?? Extremely funny considering there is NO TRUE integrated VCN platform. You order in parts through each vendor (it is not a single skew as you ellude to) and assembled in parts. Does it arrive on the datacenter floor as a single integrated stack? Nope.. Arrives in seperate components and built onsite AND managed seperately.
    * Ontap 8.0.1 release. I know you hammer home Unified and question EMC’s Unified approach but I still have to call the kettle black. Are you saying Ontap 7 and Ontap 8 are the same code base and you can upgrade from 7 to 8 with no disruption? Feel free to call me out on this one but this is what other NetApp customers have told me. To go from ONTAP 7 to ONTAP 8-C you need to wipe the disk and system (I think NetApp refers to this as a “migration” process). I would look for you to vet out the accuracy here. Is it seemless? How about going from ONTAP 8 to ONTAP 8c? Again, the customers are saying you need to wipe the disk and the system. What they did say was supported was going from ONTAP 7 to ONTAP 8-7Mode and ONTAP GX to ONTAP 8-CM is supported. You are the expert.. Would love to hear you confirm or deny. If this does appear to be true on the prior points, then it would be safe to say that NetApp is not Unified or is running multiple OS’s correct? So you mean to tell me that if I upgrade to Ontap 8-7M I can’t switch to CMode? Why the restriction?
    * Data Compression – glad you could catch up. We had that out 6 months prior.
    * OnCommand S/W Management Suite – you say EMC is good at marketing well you are doing a great job marketing this Unified suite. Let me use @MikeR analogy.. This would be that shinning bag of chips with.. what was it… a bunch of potato chip dust and hot air? Yes, I agree in this case that it is true. There was no new functionality in the release, just a packaging exercise. This was an effort to respond to EMC’s Unisphere Management Framework that we all should agree, NetApp included, is better than what NetApp has today AND IS Unified. I know you tried to drag in IONIX but that has no relevance as NetApp doesn’t play in that arena. I guess we can lob Sanscreen in that competitive arena. Has the licensing changed for that yet? It used to be licensed /port and while I was at a large IT Outsourcer but with over 6000 FC ports it was tough to pay the 60.00/port cost. I did like the interface and the beauty of the acquisition agent(talking over port 443) if you have several datacenters to run collection and analysis on. There is some cool positive stuff with the product. Hopefully it doesn’t still leverage the zoning alias’s as that often created some accuracy challenges which results in some investigation. Also their orphaned storage report is always suspect because of the way it collects the data. They orginally stated we had 300TB of orphaned storage which my investigation came down to how they were pulling data from the HiCommand server. Yes, WE at EMC are okay with giving props where due. Maybe you should try that without the BUT statement….. It feels good… Just let it out… 🙂

    So in recap, your announcement was mostly a hardware refresh of your FAS/V3100 and FAS/V6000 platforms. I can say we had a large hardware refresh as part of our announcement as well. We also introduced a kick@ss VNXe platform that is application centric with the ability to provision based on application and can scale larger than your FAS2020/40/50 product line (136 ddms vs. 240 ddms on the VNXe). We also highlighted our Isilon acquisition which you can denounce as well but denial is the first step to recovery. I guess we have to have an intervention 🙂

    Software enhancements for NetApp? None outside of the marketing on your OnCommand Suite. EMC provided a large update on the VMAX with 5875 (includes a ton of enhancements like FAST VP (which NetApp still can’t do), FLM (migrate from older Symmetrix platforms to VMAX without disruption.. Yes without disruption, which NetApp can’t do), just to name a few. For VNXe it was a complete makeover with an industry leading approach around provisioning based on the application. Can NetApp do that? Nope…

    So continue to knock our innovations and leadership as we expect this from you. Continue to discredit our acquisitions which further amazes me as NetApp has no room to question our acquisition history. Approximately 1B of sunk acquisitions is hard to look past. I am sure your shareholders appreciate that.

    You tried to divert attention to EMC acquisitions like NetWorker, Email/Disk xTender, Autostart, RM, etc but I have to admit, we are making money in those markets while NetApp is not. While you are speaking in old tongue, a lot has changed with tons of improvement. I guess this is better than making an acquisition and then disbanding it? You can try and spin the fact that NetApp is horrible at making acquisitions but it is important concern for any potential NetApp customer. I can confidentally say, this is not a worry for EMC customers. I think a bigger question is who is buying NetApp in 2011?

    Sorry for the book of a response.

  17. As a point of clarity, I do know that the SM Suite does compete in the same space as RM. I figured I would acknowledge that before you call that out. We can debate about all the competitive nuances between the products but it still goes back to Unified Management (which you can see EMC is very serious about). Single pane of glass versus the link and launch approach. I can say from your 19-20 months removed from IDS, a lot has changed including RM. You know how quickly technology changes in 7 months 🙂

    Not sure if you are a bears fan but how about them bears? Thought they might be able to pull this one off. Was hoping to see another superbowl shuffle 🙂


  18. Wow. I guess the EMC embargo is either getting ignored blatantly, or Chad lifted it.

    In any case some of the pawns are getting terribly ruffled. Breath deep.

  19. Dimitris,
    I find it interesting that very few of the points you raised have been answered or dealt with directly, though none of them are very comfortable questions for anyone from EMC to anwser today.

    There’s certainly some interesting home grown innovation in the recent EMC launch, though it wasnt nearly as interesting as I thought it was going to be. Most of the stuff which was announced had already come out and the bits that hadnt is the same stuff that partners were breifed with around a year ago. Even with a delay of almost a year from when I thought this would come out there seems to be signs of a bunch stuff that got cut out. The primary one of which is the lack of support for the ultrapoint DAE’s which if I were an EMC customer would worry me enormously from an investment protection point of view.

    It also looks like rumours of scalability issues the VNXe architecture seems to be true for the moment given its limited role in the low end. By the looks of things by the time those are worked out EMC may have something that is more or less like ONTAP 7 around about the time we’ll be delivering ONTAP 9.

    As an aside (and on principal I hate going off topic, if you’ve got a coherent argument to make about the role of aquisition strategies in increasing market share growth, write a blog of your own), NetApp’s aquisition and internal development strategy has resulted in growth rates and market share gains that our stockholders are very happy with.

    The next few years will be very interesting


  20. @Aaron – Nice rant. I didn’t see a lot of substance in there but I have to admit I started to skim once you misspelled “superiority.” (If you want to condescend, at least run a spell checker). Similar to previous EMC posts, your entry starts defensive and then predictably devolves into some NetApp ankle biting.
    Please re-read Dimitri’s original blog. The subject has to do with the EMC announcement. Your rant was the equivalent of someone asking for the time and you responding that you hate spinach. While no one would doubt your obvious distaste for spinach, it has nothing to do with the question asked. I understand the reasons why EMC would redefine itself in NetApp terms. Now that the technology is out there, I would think it’s valid to ask technical questions, right? I’m holding out hope that every EMC answer won’t be expressed as a function of NetApp. Now, without comparing yourself to NetApp (or anyone else for that matter) would it be possible to return to at least a few of the original questions in this blog? For clarity sake, how about answering #22 and #35 to pick a couple?

    @Jonas, You really don’t want to cherry-pick customer success/failure stories, do you? As you know, V-Series has been one of our fastest growing areas for the fastest growing storage vendor in the market today. Still, EMC has posted moderate growth as well. Let’s just assume that we’re both doing some things well and I would ask the same thing of you: can we return to some of the original questions posted in this blog? If you could pick a couple (#12 and #21 maybe? Your call.) we might be able to start knocking down a few of these.

    And, yes, I’ve read some of the EMC blogs on the topic and some have offered some additional detail but there’s still an awful lot of detail left to explain and it would be great if you could help the readers understand the EMC technology. Anything that didn’t start with “Oh, yeah?!” and ended with “Nanna, nanna, nanna!” would be great.

  21. Had a typo on the VNXe max drive count. I meant to say 136TB and 240TB, instead I typed ddm’s for maximum drives. Just wanted to clarify.

  22. Hi Aaron,

    Love the banter, hope you had a great vacation.

    BTW – I still am answering the EMC questions in the comments, yet nobody’s answering any of mine. The customers following this blog can see that, and they are free to ask the questions themselves – and in that situation, they’ll need to be answered. I hope you provide hard documentation when answering them, not the salesy “sure we can do that”.

    No hardware or software should be a religion, but the two top storage companies (NetApp and EMC, for anyone not paying attention) definitely have very different philosophies, which I’ll try to illustrate more clearly on a future post.

    I’ll touch upon a few of your questions, Aaron. Maybe you’ll do me the honor of answering a handful of mine. Quid pro quo?

    Data Domain: Utterly different solution to NetApp Snapshots. Neither solution solves all the same problems. DD was never meant to replace snapshots as the preferred backup mechanism, since it doesn’t do the same thing.

    For instance, Data Domain can’t back up and restore near-instantly. It is a great target-side dedupe appliance, though.

    However, for customers that don’t have NetApp storage as their primary (and absolutely refuse to change or front-end it with NetApp V-Series), Data Domain is a viable solution. Few (if any) NetApp customers already using replicated snapshots also buy Data Domain (or Avamar). It would be a colossal waste of money.

    I’ve seen replicated DD solutions that cost far more than replacing everything with NetApp… let me think, what’s best, replicate my backups only OR, for less money, get full replication of all my production AND backup data and the ability to granularly go back in time? Hmmm…

    If we’d bought Data Domain, then we would have had access to that chunk of customers that doesn’t want to go NetApp for their main disk (and, though rapidly shrinking, it’s significant enough to warrant the acquisition). Here, I gave you props and it didn’t feel funny at all 🙂

    Listen, I applaud your efforts, but the market has been speaking clearly over the past several years: NetApp grows faster than the other storage vendors. While nobody is standing still.

    Which explains (partly) why our market cap is bigger than EMC’s once you remove the VMware share from EMC’s cap.

    Not bad for, essentially, a single-product company…

    So, somehow I feel safe in thinking people will buy a ton of NetApp in 2011. Probably 50% more than they did in 2010. Which was a ton more than 2009. Etc.

    Real world efficiency and performance: Most of my customers have over 200% efficiency, and amazing performance, so I kinda think we’re good there… (there are the corner cases with 20,000% efficiency – including some companies and government agencies everyone knows).

    It somehow makes a 20% guarantee hollow. But, I know, people “bite” on the 20% EMC guarantee. Which, essentially, renders all arrays JBOD. “we’ll give you 20% more raw space than the other guys”.

    Hey – if I were all about selling cheap raw space without efficiencies, I’d be working for NexSan, selling SATAbeasts for what amounts to be Nexsan almost paying customers to buy the stuff and throwing in a free burrito and lunch-hour lap dance.

    People care about the efficiencies.

    Oh, and regarding SSD in NetApp storage: you and I both know that getting SSD in an array isn’t that hard, since the good disk companies have done most of the hard work. It’s about whether it can help or not, and by how much.

    Creating a bespoke ultra-high-performance, low-latency and amazingly efficient cache board is, however, a lot harder than throwing SSD in a shelf. But it pays for itself long-term… and it’s clever enough that it doesn’t need 5 hours to “warm up” like someone else’s cache 🙂

    SSD does help EMC’s old-school RAID architecture a whole lot, because it needed the help in many scenarios.

    SSD is a lot less necessary for NetApp, which is why we prefer cache. I haven’t recommended a single SSD so far, since I think that with NetApp’s architecture they’re largely unnecessary for 99% of customers. But, hey, we have them for that 1% that either really needs them or thinks they need them.

    On the other hand, almost every 3000-series system and above that we sell in my patch includes our Flash Cache.

    Auto-tiering: Again, EMC, HDS, IBM all benefit a lot more from an autotiering architecture than the NetApp architecture does.

    After all, virtualizing blocks of storage and putting them wherever is best is pretty much the fundamental way NetApp gear works, do you really think it’s that hard for NetApp to do autotiering at the 4KB chunk level? 🙂

    Maybe we have a better mechanism for OUR architecture. We call it the Virtual Storage Tier. And maybe it’s simpler, more efficient and totally works in real-time, unlike all other tiering approaches.

    Oh, and you can upgrade from ONTAP7 to ONTAP8 with no problems and no migrations. Unless you want to use things in ONTAP8 that need the new disk structures. Kinda like having LUNs in CX4 and wanting to use thin provisioning, FAST etc. You need to do a massive migration into pools. What about those poor suckers with CX3s? 🙂

    ONTAP8 cluster-mode is still really for scale-out NAS workloads.

    Maybe a better way to look at it is this:

    If you want to do scale-out NAS with NetApp you use the same OS running the scale-out mode and run it on the same exact gear.

    With EMC I’m looking at a completely different platform, a different set of disks, and no hope to merge the gear with Celerra.

    Are you telling me Isilon and Celerra will converge any time soon? 🙂

    Because NetApp’s philosophy is that the separation of the two OS modes in ONTAP is temporary and undesirable. Whereas EMC’s philosophy is that having multiple platforms is perfectly fine and, indeed, leads to more sales since the scale-out NAS customer will invariably also need some block storage, so they might as well buy some Vmax and Vplex and maybe DD to back them up, and some Avamar for remote sites! 🙂

    And last but not least:

    EMC has been talking about building a “NetApp killer” for many years now. Every single time, it hasn’t worked, not even a little bit. First it was the NS, then the NX4, then something else, now it’s the VNXe.

    But it’s nice that you at least recognize who you need to go after. And don’t think the customers are not paying attention.


  23. I find the musings here interesting to say the least. A friend of mine and co-worker on this thread pointed me here and said “you should post.” All ready to roll up my sleeves and throw it my best jabs, something dawned on me and I decided to keep my post short (mostly just to keep my promise to my good friend and fellow EMC-er).

    That something: in the end, it simply doesn’t matter whose buzzers ring louder, blink faster, or whose propellers spin longer. What matters is can technology solve business problems? Can a vendor present an ecosystem of solutions to businesses to address their needs and help them achieve their goals (enabling their business)? The answers to these questions don’t have anything to do with the technology itself, but how that technology can be applied to create solutions. This idea, this concept, is what to me is so important about the organization that EMC has become and continues to evolve into. We have a portfolio of solutions that enable us to have a variety of conversations with businesses over a large breadth of technology domains. This, above anything in any announcement, is the core of the strength of EMC. And, quite frankly, puts in a place that a single track organization like NetApp can’t follow. Contrary to your statement, we exemplify though leadership (I’m sorry, where is your FAST?) in not only the storage space but the overall infrastructure information space.

    Dmitiris, honestly, do you really think a competitor is going to explain the DNA of their new product to you on a blog post? Or, perhaps more to the point, who are you to ask such a thing? I think if you really thought about this, you’d realize that if you want the answers to these questions you must obtain them yourself. And I don’t see that as shifty or evasive on our part, I see it as common sense.

    And lastly, Dmitiris, remember that this discussion about mid-range storage is the only discussion that our companies can have competitive dialogue about. Frankly, there are so many other discussions that EMC is having that NetApp simply can’t be a part of (despite wanting the FAS, ONTAP, etc. platform to be the technology industry’s Swiss Army knife).



    P.s. Aaron S. did a, IMHO, great job summarizing your companies recent refresh announcement. And, from what I have seen I could argue it looks like a minor refresh of current paradigms. But I won’t, but I’m not too interested in that. I have only one question for you, how would you use that platform to store, protect (backup, replicate, etc.), secure, virtualize, content manage, data warehouse, or archive? Or perhaps better, can you do all of that? I know what our answer would be … absolutely!

  24. Jason,

    Thanks for your post.

    I’m not too interested in hardware refreshes from anyone (including NetApp). Software is where it’s all at. FLARE30, for instance, was released months ago. There’s nothing materially new software-wise in the VNX. VNXe is more interesting for EMC’s future, even though it’s lacking a ton of features now, and in essence it’s been out in NX3e guise for like a year already, just not worldwide. So that’s not new either! Enginuity 5875 for the Symm also wasn’t new (for the record, I do respect that release immensely).

    My whole point was that there’s no new DNA to explain anyway, aside from SAS (last vendor to do it?) and faster CPUs/more RAM.

    The software is the same, but the packaging/licensing is different (to better mimic NetApp’s).

    I actually thought the Data Domain announcement was the most interesting part of the whole launch 🙂

    Read my previous response Re FAST.

    I firmly believe ONTAP can solve about 80% of storage concerns and, consequently, grab that percentage of marketshare. The other 20% either is too niche, or NetApp has no interest in pursuing since we are extremely focused. You will see the percentage of tasks ONTAP can deal with grow beyond 80% in the near future. Maybe by more than you realize.

    That doesn’t make the 20% worthless, and EMC and everyone else is more than welcome to play there. It’s a big market.

    After all, the world doesn’t need yet another Documentum or Filenet. We’ll be happy to provide highly efficient and resilient storage for both.


  25. @MikeR – I should submit that your one previous response had substance? It reminds me of my days playing arena football and you had that “one guy” in your opponents section who did all of the trash talking but he never suited up. Go figure right…

    I appreciate your heads up comment, sarcastic or not, regarding spelling errors. It is not uncommon to have spelling errors on a blog but I will make a better effort to satisfy your request. Any spelling errors at this point, MS didn’t catch. Let me apologize now if there are any.

    No comments regarding details of NetApp’s launch? Or is it not up for debate? If yours was so impressive, should be just as easy right? It was just a technology refresh… period… Don’t waste your breath.

    Look, I am not a bad guy and I want to be in your good graces so let’s answer one or two for ya big guy….

    14.Why is FAST still at the grossly inefficient 1GB chunk?
    A: FAST is not specific to the VNX or VNXe launch. One might suggest like many other efforts, EMC will seek to improve things further with each new iteration of the microcode, as EMC always does.. I would like to see this become sparser but I have no control there. I was not aware NetApp had a FAST Like capability to automagically tier from SSD, FC, and SATA? Please do tell…..

    9.Can FAST on the VNX exclude certain time periods that can confuse the algorithms, like when backups occur?
    A: Without diving into the algorithm, the answer is yes. As you point out, nothing worse than promoting a hot chunk of data that is not worthy of promotion e.g. backup. There are ways to pin data within certain tiers if needed but in your case you are referring to backup. We can analyze data over time and can determine the frequency of access to the data. So, it would be pretty easy to isolate backup data based on the frequency of access. If the data is only accessed every night between let’s say 10pm-4am, we can easily identify this as a “false-positive”. This is my term not EMC’s. You can set days and periods of time such that there are “exclusions”. For those dynamic workloads, this is where FlashCache and FAST Cache come into play.

    @Dmitri – Thanks for your responses. Just saw them.

    I didn’t see any comment on your VCN or vBlock equivalent? Can you provide the no marketing details? Is it really one skew? Does it come pre-packaged to the DataCenter? Is it ONE single management framework? It is marketed like it is this seamless stack and I just haven’t seen that. I am also not getting this same feedback from NetApp customers either.

    Can you walk me through the process to “migrate” from Ontap 7 mode to Ontap 8-7mode? Also from Ontap 8-7mode should be a quick and easy to Cmode right? You are already to OnTap 8-7mode so this should be a simple flip of the switch right? I think you referenced it like migrating a thick lun into a pool (CX4). I know of several ways to complete that progress which isn’t anywhere close to the adjective you used “massive”. Plus your customers probably want to know what this simple yet potentially cumbersome migration means right?

    I didn’t say DD would replace the NetApp snapshot philosophy but it is very interesting that now DD is no longer a play, snapshots for backup is the right answer? Just like there is no indexing with snap shots and when customers want that, you quickly shove in Syncsort (we know NetApp doesn’t like to push Syncsort) because that is all you have (we won’t talk about the limited focus with that product). I remember those days where you heard about snaps in almost every campaign for backups, then that drizzled off for a while when the DD bidding took place, and then it was back to “Yep, snap, snap, snippity snap”. I can confidently say NetApp acquiring DD was not just for the customer base (nice added bonus however). It is because NetApp didn’t have a disk target, is a hot product, and this would, “compliment” your line-up (yes, I used this word). The very same word you beat us up over for having a portfolio. I am confident you would not be saying snap, snap, snippity snap, snap for everything. If you did, it would cannibalize your DD sales. I am buying a shift in messaging more than anything.

    I am curious what customers you service because DD and Avamar are in a ton of NetApp accounts that I call on. Just like NetApp is in a lot of EMC accounts. We both have our loyal customers and then there is a lot up for grabs.

    Let’s save the whole NetApp snapshot versus Avamar or DD discussion. That is a longer topic with pro’s/con’s for each. NetApp is not this peachy solution for every use case which you might lead everyone to believe.

    We should probably be careful when we talk about “warming-up” as NetApp is not exempt from this especially in a failover scenario. Let’s just say you lose filer head A which means a “takeover” occurs and you lose 50% of your MEGA Cache which will require a warm up period. Conversely in an EMC world, losing an SP does not require a “warm up” period because it is a “shared” FAST Cache architecture. Are you stating there is no “warm up” period with FLASH CACHE and also no warm up in a controller failure?

    Those guys with CX3’s can upgrade to CX4’s (DIP) and we can migrate their devices into Pools. CX3 customers have a path and not sure where the “massive” migration comes from. Again, I would reference what it takes to go from OnTAP 7.3 to OnTAP 8.0 Cluster mode or OnTAP 7.3 to GX. I am aware you can go from 7.3 to OnTAP 8.0 7-mode through NDU and GX to 8.0 Cluster mode through DIP. We have to cover all of your customer base running these different versions.

    To expand on this further, what if a customer use case is more resiliency than a 2 node architecture can provide. I hope NetApp’s response (since you have thousands of OnTAP 8.0 customers) is “we can help”. What if a customer’s protocol of choice is a hybrid approach of FC and maybe CIFS/NFS. This isn’t just a scale-out NAS approach like what you mentioin. So I want to go from my current 7.3 to OnTAP 8.0 Cluster mode. What is the migration process? I am confident the answer is “wipe-clean and recopy of the data”. Are you saying this is no longer an issue?

    You can run scale out on the same gear but it is a “massive” effort to get you there. I am curious about that detail. How much work is it? Spill the beans Dmitri 🙂

    I guess I don’t look at multiple platforms in terms of more sales, rather, I look at the ability to address more customer challenges by leveraging “options” vs. the “one-way” approach. Depending upon the customer requirements, they might require a VMax, Vplex (more scalable than the VSeries btw..), and DD if needed. Who knows if it is a remote office, you can always say “let’s leverage OSSV” but are we really sending less data across the wire? No, you send everything because OSSV to my knowledge doesn’t do the hashing through the agent. May have changed but not what I have been told. So, it is an “it depends” and depending upon your business challenges, environment, etc we can provide a solution. I tend to build a solution for the customer rather than forcing the customer into a technology choice because that is all you have.

    Not sure about the “NetApp Killer” comment but I can confidently say, we can hold our own 🙂 I have no issues fighting the fight and providing the right technology for the customer. We surely wouldn’t go from an NS to an NX4 and then back to the VNXe. That is like saying the FAS2000 series will address an enterprise customer challenge based on their size and performance requirements. NX4/VNXe address a different market segment compared to a VNX (this kind of answers another one of your questions) It is no different with NetApp. You are not going to propose a 6000 series to an SMB customer right? No, you are going to hopefully collect all of the necessary data to properly size the solution based on customer requirements.
    I don’t expect to convert you or any of the NetApp fanbois. It is a religion like I said. There is only one option at that is NetApp in your world. I guess I can’t say any differently on the EMC side except, we have more to offer the customer to address those challenges within the four walls of that datacenter.
    Look forward to the continued banter. You won’t get either of us to agree that one technology is better than the other because we are passionate about the technology we sell. We believe in it. Not a bad thing I guess. Have a good night. More tomorrow…….

    1. Hi Aaron, I know the FlashCache contents can be maintained after controller reboots so long as it’s a clean shutdown (i.e., takeover/giveback or reboot). Is that what you mean, or are you asking whether or not controller A can “take over” controller B’s FlashCache?

  26. @Dmitri

    Read your last post and you made me laugh… I love the confidence (I might question whether that is confidence or another adjective), Boy are we two very similar competitive cats. 80% of the market is a nice yet lofty goal 🙂 I set some pretty lofty goals as well. Let’s just say, you will have a lot of work cut out for you. Folks like myself might have something to say about that 😉

    I have to also stress in NetApp’s Mega Launch there was nothing that materialized from a HW or SW standpoint. NetApp can chastise EMC all you want over the launch but I am still not seeing anything impressive from your November launch. Sorry… Please fill me in on what was so earth shattering.


    “My whole point was that there’s no new DNA to explain anyway, aside from SAS (last vendor to do it?) and faster CPUs/more RAM.”

    *Maybe you can explain why NetApp was the last vendor to integrate SSD technology since it is so easy to integrate (according to you) or why they are the last to offer a FAST-Like offering? You should be able to whip up some FM FAST technology beings that it is so simple and you can promote/demote at a 4KB level with zero overhead right?

  27. Q. Why did we offer SSD’s later than everyone else ?
    A. Because Flash as cache is a better approach that adds higher value, so we concentrated on doing this first. SSD’s by themselves satisfy a very small number of workloads, especially on a FAS array whic accelerates write workloads by their very nature. A better question would be, why did it take EMC and everyone else so long to emulate this approach

    Q. Why are we the last to / why dont we .. do a FAST-Like offering ?
    A. Because Virtual Storage Tiering (which is fully autmated) via intelligent caching is a generally superior offeringif you do it right, especially when combined with the other netapp technologies as it adapts in real time, unlike FAST. It’s interesting to me that Virtual Storage tiering in association with FlashCache was available for almost a year before EMC finally released FAST V2 (and we all know what a horrible kludge FAST V1 was)

    By concentrating on the approaches that add the most value, and providing them to our customers before the rest of the industry, NetApp ensures that the majority of our customers are able get value early rather than providing technology that merely looks good on a marketing brochure.


  28. I guess this thread proves one thing. EMC’s product launch doesnt seem to be as cool as putting people in a car. It seem the industry has picked up that its nothing new other then some Data Domain Stuff.

    I think it raises more questions then answers as part of the launch.

    New CPU and Hardware is good, but its not EMC’s biggest ever launch as they advertised

    Thanks for all the great commentary


  29. Whoever first made the joke that with a suitable dedupe technology, the number of EMC racks jumped would have been reduced by 90%, was a very funny person.

  30. Aaron,

    I’ll have a large post regarding Vblock, that will take care of your questions there, up later today or tomorrow. Too big to put in as a response.

    Oh, regarding warm-up of cache: You frame your concern from an EMC viewpoint. I saw the SQL benchmarks you posted with FAST Cache – it took over 5 hours to warm 140GB of cache! Naturally you want it persistent if it takes THAT long. Imagine on a fully loaded CX with 2TB cache.

    NetApp Flash Cache warms up way faster (a few minutes) even though it’s so much larger (the minimum in a system is 512GB) so even in the event of a failover it’s not such a big deal.

    And it’s funny because that’s the biggest knock EMC has on our caching.

    You know what, I’d rather have:

    1. A far lower latency cache, directly on the bus, running on a custom board with crazy high performance and not generic SSDs
    2. High efficiency and granularity (4KB cached vs 64K minimum on a CX) and
    3. Tunable cache that can even accommodate sequential workloads
    4. Extremely quick warm-up
    5. Deduped cache

    Regarding migrations – ONTAP 7 to 8 is a straight upgrade, no downtime. Flipping the switch to go from 7-mode to c-mode: today, not seamless. Of course, Celerra has no c-mode equivalent really, so there’s no comparison there. Will you enable seamless switching from Celerra to Isilon I wonder? Ever?

    Anyway, if you go through the comments it looks like NetApp answered 15 questions, and EMC maybe 3, and I think you’re missing the point a bit:

    The point is that, when comparing ONTAP 7-mode to DART/FLARE, the DART/FLARE combo is severely lacking in its unification and features (because that’s the most appropriate comparison, we can compare Isilon to c-mode another day – and, funnily enough, EMC was badmouthing Isilon up until a few months ago, now it’s the bee’s knees – and the amount of badmouthing regarding Data Domain was nothing short of staggering. You’ll be happy to know that EMC was upset with me because I refused to sell any more of the crappy DL3Ds and instead sold Data Domain, I was just ahead of the curve).

    Combined with the lack of support for older EMC hardware (the new VNX systems flat out don’t support the old disk shelves) means that, for existing customers, there’s zero investment protection.

    All older NetApp systems with 64-bit CPUs can take ONTAP 8.01 and take the old disks with them, and also add SAS shelves. No need to replace stuff. We’re talking 30xx, not 3100, so 2 gens back.

    To even get FLARE30 you’re talking controller upgrades, even if one had a chunky CX3.

    It’s all part of the questions you didn’t answer 🙂

    It’s cool, I know it must be difficult for you too, since you must have customers that have purchased a bunch of FC DAEs and now have to chuck all that if they want VNX. It’s the price of progress.

    We will be more than happy to virtualize the storage of those customers and give them advanced functionality with NetApp V-Series 🙂


  31. Funny that Shea thinks we have an “embargo on banter” headed up by Chad as some sort of “blogging mob boss” :-). You guys clearly don’t know Chad and quite frankly, even if we wanted to do such a thing, we aren’t close to being that organized. I think we do feel sometimes, because we are #1, and get attacked by everyone who wants our market share and getting into it at all is like laying down with the dogs.. I’m fairly confident at this stage it’s only “Us”, that is, EMC and NetApp people reading this blog, so let’s take it up a level for a second:

    Isn’t it stupid when when one of us (“Industry leaders”) does something like this when it reveals many misinformed and outdated assumptions? You guys will never have the expertise we have with EMC (regardless of how hard you may try)..which is why it’s so easy to be wrong when you talk about us. We don’t claim to be experts in your stuff either btw. We both have many people who worked in each others companies at one time but there is always a “rate of decay” that rapidly sets in compounded by the fact that things tend to change so darn fast, man..its hard to keep up with your own company’s tech, let along everyone elses. By the time you circulate the “competitive information” and it finally saturates your field, it’s likely wrong already. I can’t help but wonder how this kind of stuff helps customers who already have trusted their advisors and account teams in place exactly?

    Now ask yourselves if an EMC employee posted a list of 35 questions about NetApp that showed a certain degree (understandable) ignorance, outdated\stale content and clear malice (that is only really ultimately about developing a ntap/emc compete strategy), how does any of that help the customers we serve? If there are real customers out there who see value in this type of back and forth, directly for themselves, then I’d love to hear from you either right here or directly. What most of you customers have told me is you don’t find this kind of rubbish useful. If you are out there, I can assure you all (customers) we (EMC ) will take complete care of your needs and answer all YOUR questions as they relate to what it is you are trying to do. BTW – please neither emc or ntap people “manufacture a customer” to magically appear on the post as that would undermine and make a mockery of who ultimately pays all our bills (the customer).

    We appreciate the list of questions as a merely an “open relay” broadcasting your intended competitive selling motion. By placing it in such an open forum like this Dimitris, you make us better because we have time to formulate a well thought out, strategic response to circumvent the intended rat holes you’d like us to go down. The answers to what you’re looking for are almost done, and I’m sure you will end up with a copy in hand at some point in the future (through the usual avenues that aren’t *supposed* to exist but we all know they do). If I feel like it, I might come back and answer your questions (when we are done).. The one thing I keep coming back to however is that each question will branch/morph into many other questions..and again, how does this help customers exactly? M Perhaps we need to barter around the questions to some degree..I might play a bit more and pass a few questions to you, if I think it’s productive and beneficial for customers first and foremost. I’m open to hear from customers first and if this is the best avenue (a relatively unknown blog in the IT industry) to get the information to them you will hear much more from me shortly.


  32. Hi Jonas,

    I’m sure some are wrong, which is why they’re questions and not statements. Indeed, I will edit a couple after I post this reply, since I now know the VNXe doesn’t use a hypervisor to run FLARE and DART, for example. Hey, it’s almost a chance to say “yes, this was a problem before but we fixed it!” – which is progress.

    I won’t comment on the embargo stuff since I have no idea – the only embargo that makes sense and that people really follow is to not reveal info on products before release.

    In the field, we encounter incorrect documents from various vendors all the time. You know what, I’d almost welcome the chance to fix the incorrect bits, just so we can get on with our lives and not have to fight the same FUD over and over again (I offered that to Chuck once in a private twitter exchange). The problem is, there will always be the rogue sales elements that don’t care and just want to sell the stuff since if they don’t make their number they’ll get fired, so they’ll do “whatever it takes”.

    Indeed, let’s try to solve a customer’s problems without having to fight competitor FUD. And may the best and most cost-efficient problem solver win.

    In my experience, in the field, EMC is a prime source of such FUD, HP comes second, Dell third, but I’ve seen some choice documents from smaller vendors, and blogged about them in the past ( and for example). So, the holier-than-thou attitude is a little bit weird.

    I agree some technology changes dramatically even in a short timeframe, but some just – doesn’t. For instance, the dedupe size for a NetApp volume has been 16TB for some time now (and you can have many of those). It’s no secret and no big news. Celerra volume sizes are also at 16TB and have been for a while. NetApp can go to 100TB (relatively new) but loses dedupe.

    Other stuff changes far more rapidly (for example, performance optimizations).

    The intent behind a lot of the questions is to expose the fact that, no matter how much one dismisses it as unimportant, ARCHITECTURE MATTERS.

    Before EMC bought DD, the party line was “dedupe is just a feature”. Now it’s one of the hottest products EMC sells.

    Because, let’s face it, its architecture was way better than the DL3D. You see, both products tried to do the exact same thing, only the execution on the DL3D wasn’t quite as good as the DD one. It still was easy to use (so it wasn’t a GUI thing), but unless you configured the guts of it just right, there were performance issues and more. And if something went wrong with the guts (that relied on Stornext, a dedup engine and some VTL code – several separate bits), fixing it was, charitably, tough. But it was the best EMC could do at the time, so they stuck with it.

    So, the moment EMC could, they bought the far superior DD and kicked the DL3D to the curb like it deserved. Slootman even famously swapped a bunch of DL3Ds for free.

    But, until then, it was all about how to beat DD with DL3D, partner SPIFs, tons of competitive documentation trying to make DD seem bad, and (what bothered me the most) partner arm-twisting (“sell DL3D or else!”)

    This is not second-hand knowledge to me, I was one of those partners involved with the DL3D getting my arm twisted and refusing to sell more after some VERY bad experiences. EMC management in Chicago still hates me because of that. Because they felt that, on principle, I should sell EMC products no matter what, even though I had the choice to do the right thing for my customer.

    We all have a living to make, and, sometimes, what we sell isn’t ideal. There’s no perfect product. Otherwise, everyone would be buying that one and nobody else would be selling their stuff.

    I do believe in customer education, and having the customer ask the right questions, as tough as they might be. It makes us all better. Let’s face it, EMC and NetApp need each other, since without competition, we become complacent. In the end, the customer benefits.

    Ultimately, it’s all about solving problems efficiently, it’s not a religion and there are 20 different ways to do the same thing.

    A customer with a finite budget, personnel and time will choose differently than another customer with a huge budget, no personnel issues and a year to get things done.

    And, you know what, some customers will be OK with the EMC architecture in the NS/VNX, warts and all, others will prefer the NetApp way because it’s better for them, warts and all.

    But I still want customers armed with the right questions, whether you answer them in this unknown industry blog or not.

    Sorry for the tirade, the wife made me goulash and I can barely think straight… 🙂


  33. I’ve Read 37 comments, and i still did not see a single straight answer to Dimitris’ questions.

    To the EMC crew…If you have NO intention on answering a question then don’t freaking post.

    To the guys answering a question by writing an essay…This isn’t an English Composition class. Stop trying to divert attention to other irrelevant matters and concentrate on the subject of this this post. Stop the posturing bullshit.

  34. Hi Nick, to be fair it’s (by intent) an unmoderated blog, not a corporate one, and people can post whatever they want. Actually I didn’t expect any responses from EMC at all, it was more directed at customers so they know what to ask for.

    If the answers they get from EMC are to their liking, then so be it.

    But at least they’re aware of certain things to ask. Even if the list of questions preps EMC. It’s not as if they’re suddenly uncovered issues, I’d expect EMC to already know how to address all of them in the field (and hopefully customers will expect more than “not an issue” as an answer).

    I know the trend is to make storage systems a black box (see XIV) but my whole point is that architecture matters, and just calling it unified doesn’t make it so.


  35. @Nick

    Nice class, and quite typical in the NetApp blog o sphere. We have to find out all of the competitive details through our own efforts just like you. What we focus on is what matters in the “customers eyes” not NetApp’s attempt to derail customers into rat hole conversations that solve zero customer issues.

    For the record, I was just keeping up with Dimitri’s essays 🙂 Speaking of posturing, what was so earth shattering in the NetApp’s launch Nov 9th? No takers? You know all of those innovative HW/SW enhancements? If you want to condescend on our launch I am merely asking “Where’s the Beef” in your launch?

    Yet another guy that stands in the back of the crowd and says “Yeah, what he said”.

    Dimitris – Well said. We at EMC believe Architecture is key as well so that isn’t in dispute. Neither NetApp or EMC would be in leadership positions if your HW/SW stack wasn’t considered superior in the marketplace.

    How was the goulash? 🙂 If you are like me, you are miserable by the time you get done with eating goulash. Never can eat just one helping…

  36. @Aaron: 3 helpings of goulash. Was dizzy after.

    Regarding the NetApp launch:

    The typical performance/expandability hardware refresh (with the high end getting a big bump), the new ONTAP features were more interesting with fully unified connect (use 1 cable to transfer all protocols – something others still can’t do in their arrays since, duh, they’re NOT UNIFIED so it’s impossible), performance enhancements, compression and more. So, very good software enhancements that materially affect the solving of certain problems, including vastly simplified cabling and switching, which for some reason everyone but the poor bastards dealing with the cables seems to ignore.

    Do I think it was an earth-shattering event? Dude, I am so jaded that unless a company manages to resurrect Gandhi, cure cancer or AIDS or somesuch (ideally all at the same time), I don’t think it’s a huge deal.

    The stuff from both EMC and NetApp launches were all mostly incremental improvements, especially in the light of FLARE30, VNXe (née NX3e for a YEAR) and Enginuity 5875 being around well before the launch. To me, the only truly new thing in the launch was the Data Domain stuff 🙂

    So, IMO the EMC launch was more like a comprehensive summary of stuff that was ALREADY OUT, with the emphasis on Bubba jumping with his bike and a troupe of dancers being stuffed into a mini (which must be a human rights violation in some part of the world 🙂 )

    But anyway, it’s not important, what’s important is whether the new boxes solve customer problems materially better.

    My point is – aside from the faster performance due to new hardware and the new wizards in VNXe, I don’t think there’s a material advancement in how they solve problems since they’re just a rehash of existing stuff. I’m not saying they don’t solve problems, I just don’t see the new, improved ways in which the problems are solved.

    Time for more goulash!


  37. Your company wouldn’t know class if it bit it in the arse. I mean that.

    Explain to me, how classy is it to spray paint your competitors sidewalks on the day of YOYU launch!!!! YOUR LAUNCH.

    Now, that’s…classy. maybe you want a picture?

    As far as posturing we know your company’s diversion strategy well… Write an long essay in the comments section of a blog in an attempt to confuse readers and divert their attention to irrelevant matters or subjects that suit your better.

    Answer the questions and stop banter

  38. @nick

    or all of NetApp for that matter.. I speak for the new, younger generation of EMC where we work hard solving all sorts of customer challenges and a byproduct is obviously the competitive nature of the business but also having FUN in the process.

    Since I wasn’t aware of this (honestly), I am now going to hop on the google and check out what you are referring to.

    Initial thoughts, all in FUN bud. No harm or disrespect, just two companies that have very competitive spirits.

    My guess would be it wasn’t spray paint as that just wouldnt come off that easy. Dust it off and don’t take offense. Look at the NetApp Dedup video on youtube puting “DD and EMC on blast yo!”. All in fun, it was quite humorous and we all liked it.

    Don’t take the job so serious. Take the customers serious and have fun man. You should have fun with NetApp being in the top 100 (Top 10 I believe – Congrats BTW) for best places to work. You would think people wouldn’t be so stressed out. Sheesh…

    Lastly, sometimes you can’t just answer a question with a yes or no because of the context in how it was asked.
    For Example: Customer says I want EMC, are you going to say — Sure? You might say “well, I am from NetApp first of all and I would like to talk about how we/NetApp can solve your business challenges” right? I hope so, or don’t let Dimitri know about it! Lord knows he has something to say about it 🙂

    I am merely joking here so don’t go crazy on me now. NetApp is a great company with lots of great people working there. I know, I have many friends there. Does that mean we can’t walk across the DMZ and shake hands, have a beer, or go play dodge ball? No! It just means, you can drain the NetApp Pitcher of Kool-aid and have fun with the “other” Kool-guys.

    And my final note, I represent myself and EMC with full transparency, honesty, and integrity to OUR customers but I have no obligation to be transparent to you/NetApp. This does not mean I don’t respect NetApp because I do.

    How about a hug? I am just a big old teddy bear.. J/K.

    Good selling!

  39. That looks like a stencil on top of cement that was pressure washed? If so, we cleaned NetApp’s sidewalks at no additional charge 🙂

    Seriously, this was purely meant for fun and shouldn’t be considered any more. No defacement of property as it doesn’t look like spray paint.

    I am sure we could get upset about NetApp’s video on youtube where Matt Watt’s talks about a resource within NetApp (I won’t name the person) that was using an EMC only tool (Cap2,capacity calculator,etc) in an effort to come up with some bogus capacity. That is probably more of a legal issue than anything. I wouldn’t publicize that honestly..

    All in all, sorry if you guys took offense to the spray wash of the sidewalks. All in the competitive spirit.

  40. Predictable response Nick.

    Still no customers as I suspected which makes this entire thing..a sort of futile, vendor/competitive masturbation exercise.. so I’m least for now.

  41. Disclosure – EMCer here.

    Holy cow – this comment stream is awesomesauce 🙂

    No better example than vendor navel gazing will EVER exist, and Dimitri, you must save it for all time 🙂

    Looking at the comments, I cant help but notice… it’s all us talking to us 🙂

    Personally, Im here in Barcelona (have you ever been here – beautiful place!), having a ball, working with the vSpecialist team here in EMEA.

    We just went out on the street and held out a homemade “hugs free” (“abrazos gratis”) sign. The hugs we got from strangers were insanely gratifying. Consider that highly recommended (to huggers and huggies.

    That was after a great day of great business and an evening of go-carting (which was physiologically much harder than I anticipated).

    ONE correction – seriously guys, if you think that I have some sort of “blogger embargo”, you ascribe FAR more power to me than I have (it sounds awesome, sign me up), or a conspiracy where there is none… and you know what they say when you start to see conspiracies everwhere 🙂

    I have said, and will say again – often I think that the blogosphere is far too much insider ca-ca (take that! :-), and far too little stuff that helps customers. I’ve also said internally that EMC, as the storage leader, responding to baiting only helps our friends in Sunnyvale. I’ll say it publicly. Hey, I just did 🙂

    If people want to listen, great. If they don’t – great.

    I’ll say one thing more (that I’ve said publicly on the record several times) – people should give LESS THAN ZERO credence to what a vendor says about the other guy. In general, they are FULL of it. (and that’s coming from a guy who has his own NetApp, EqualLogic and other kit in the home lab – just to keep sharp, and think of the realm of the possible).

    In fact, if I was customer, I would actively put vendors that bash the other guy in the penalty box (some nice Canadiana for ya).

    Those perceived slights are in the customer eyes, not in ours. We’ve CLEARLY demonstrated that our collective skins are, shall we say, a wee bit thin on the topic of perceived slights (me included)?

    Customers should give only slightly less than zero credence to what a vendor says about themselves. They should give the greatest credence to what their peers say, and the MOST credence to their own experience. Judging by that metric, EMC and NetApp are doing just fine, and are fine companies, filled with fine people, with fine innovation and fine products.

    There’s no place for fanatacism in our business – no? Shouldnt we be pragmatists?

    Let’s all take a collective chill pill, shall we?

    Have a great day!

  42. Hi Chad, thanks for jumping in. Hope you’re having a great time in Barcelona.

    You’re one of the more reasonable voices in the blogosphere, but vendors will say stuff about vendors – it’s too easy for the salesfolk. Your attitude seems to not be prevalent in the field, at least where I play.

    Sometimes the allegations will be correct, sometimes not. Sometimes there’s malicious intent, sometimes not.

    Just today a customer told me HDS was telling them to not go with the 450GB drives from NetApp because they will be end-of-lifed. HDS, of all people – telling my customer that the 450GB drives I provide will be end-of-lifed, so the customer calls me in a panic asking me if it’s true and should he get the 300GB ones instead?

    Weird thing to say for the most common 15K drive type in the market. I had a “WTF kind of competitive strategy is this” moment. And I will keep calling them out as they come if I have the time. Maybe I’m the self-appointed Robin Hood of storage.

    Seriously, maybe it’s my patch, but I see the craziest things all the time.

    Some of them come from you guys, I still see the silly documents stating NetApp has like 30% efficiency (which doesn’t even make sense, nobody would buy the stuff). Maybe the field needs updated documents, but something tells me they’ll ignore them since the old ones have more “punch”, even if the end result is usually not the one intended.

    I think I’ll start asking customers to put the vendors they care about in the same room and let the fun begin. Any childish antics and machismo will be squashed quickly, and it will be a far shorter sales cycle.

    I’ve done it a couple of times and it was actually quite civilized (once with Dell in the room, once with an EMC VAR).

    Otherwise there’s just too much FUD back-and-forth, everyone spends too many cycles, and the problems stay too long unsolved.

    I’ll even hug you and Aaron provided there’s no monkey business! 🙂


  43. @Aaron,

    Washable or not is besides the point. The important point is indicative of a poisonous, mean spirited, a completely classless culture with much disrespect for its competitors.

    @ Jonas,

    If you care to make comment with substance on this, then i’m all ears…

  44. Chad –

    Seriously? You did the “Free Hugs” thing?? Very cool. That officially makes you part of the 1% club of people willing to do completely new (and uncomfortable) things and learn. Grow, you human, grow!

    On the embargo – I heard it from too many EMC’ers. Perhaps some gaming going on – but i doubt it, they do not all know one another.

    But what ever the case may be – I’d have taken you up on the sign if I had been in Barcelona! And yes, it is a magnificent place. Try Mallorca next. Excellent.

  45. @ Mike, Nick – I’ll always have a hug for you (or anyone) – the most awesome hugs came from the people you least expect.

    It wasn’t me who started it (I can’t take credit), it was a dude Fred Nix on the vSpecialist crew, but it was too awesome to pass up.

    Re: the “embargo”….

    If someone told me to not say something that needed to be said (IMO) on my blog, well, I’d say it, and dammit, if I get canned… well, I get canned.

    There is **NO** master plan when it comes to EMC bloggers. There is a blogger DL where sometimes we talk, but that ain’t a conspiracy. That’s people talking.

    Note that we DON’T have our blogs under some * domain. something BTW that I vehemently disagreed with up front (thinking that a coordinated professional model would be better), but wiser folks pointed out to me that I was wrong. And I was (and will be again).

    The ONLY rules are:

    1) disclose who you are always (we have a standard disclaimer we need to put in our “about” if we author a blog, and are asked to always disclose who we are when commenting in other places)
    2) don’t violate EMC’s code of conduct (just like any employee needs to consider at any company)
    3) don’t disclose EMC confidential information (same as any employee needs to to consider at any company).

    That’s pretty well IT.

    There was of course a conscious effort to not talk about stuff prior to the Mega Launch, but, hey that falls into “don’t talk about unreleased stuff”.

    The only other things that could POSSIBLY be construed as an “embargo” was my continued guidance to not get sucked into these things with you guys (or anyone else). Of course, here I am… sucked in 🙂

    When it comes to all sorts of things…

    I will say this – what people think is a “well founded” or “mean spirited” observation/comment is very much in the eye of the beholder. Believe me, I have a million examples in the other direction. In YOUR eyes, you would look at them and say “seriously?”, and I would say “SERIOUSLY.” As ridiculous as you find the 450GB HDS story, or the EMC competitive team’s comments to Netapp around utilization – THAT’S how funny your original list at the top of this post seems to me. SERIOUSLY. Pause and think about that. “correctness” and “intent” are very rarely black and white – are often much more subjective.

    It’s mind blowing (it’s mind-blowing if I try to imagine the tables turned). As baloney as you think the examples you point out (the HDS one, the utilization one) are, I think the same of the list you posted. I can spend time arguing with you, but as we’ve discussed, this is a giant navel gazing exercise.

    And hey, if arguing makes you feel like the “robin hood” of the storage world – all the power to you. Of course, I don’t think it’s noble to post a list of FUD that’s off in so many ways I don’t want to even count, but that’s the nature of different world views. You view it as being a “bearer of the TRUTH”. So does that HDS dude you pointed out that stupidly thought he knew your products better than YOU.

    I’ve done the “cage match” thing (multiple vendors in the same room), and will do it anytime, anyplace – and it’s ridiculously fun, but at the same time, kinda silly too.

    It’s OK. I stick to my proverbial guns on fundamental points (of course as a Canadian, the “gun” analogy doesn’t work, and I consciously recognize that I fall off my high horse too):

    1) that when we attack each other, rather than focusing on the customer, and what we can do to help them – it doesn’t help, and only brings us down. It’s likely inevitable, so hey, Chad/Nick, get over it and deal 🙂

    Personally I think there’s a realm of “competitive fun” that is OK.

    Also, personally, I think the “NetApp rap” that trashed EMC and others was in good humor and amusing. It DID make me angry, but at the same time, I though “that’s kinda funny”. Likewise, personally, I think the power-washing was in the same vein. Obviously we’re not going to agree, so I’m not trying to argue the point (let’s not waste our binary breath). To me washable vs. “acid etched” is a world of difference, but hey, maybe we disagree on that too.

    2) When things get personal (man, look at the latest diatribe by Tony Pearson), it’s just not cool.

    Here’s to finding an opportunity to give each other a hug 🙂


    I don’t want me to have the last word – please respond 🙂

  46. Hi Chad,

    I’m still curious which of the questions are inappropriate/inaccurate. The more I check the information, the more it points to the fact that most of them are still accurate (some of them I really don’t know about, which is why they’re questions – they at least used to be issues in the past).

    I mean, seriously – if EMC has actually tackled all 35 concerns, then that’s a huge deal and something to be celebrated, and I might as well go work there. But when I see certain things like the 2TB limit for cache, how is that wrong? Are you saying a VNX7500 can go to 4TB cache? 8? 16? More?

    Or Recoverpoint being able to restore a celerra pool as one by putting it in a consistency group, effectively keeping it unaware of the Celerra structures. Good if you want to roll everything back I guess. NetApp has this, we call it the Aggregate Snapshot, and you can roll back entire pools. It’s been there forever, and can save your bacon, but we don’t advertise it as a major capability of the system. So, marketing kudos, but this fact about Recoverpoint further emphasizes the dichotomy between old-school FLARE and DART in the VNX.

    So, unsure how I’m wrong on that, either.

    Or the deal with FLARE pool striping:

    I could go on. I also want to clarify that practically all the questions are aimed at the VNX, not the VNXe.

    So, I think I’ve probably got a handful wrong, but that handful really irked you guys, and I’ll be happy to fix them if told I’m wrong. I’ve done that before for other posts, for instance this one: – I should probably add VPLEX there.

    Regarding the stenciling of NetApp sidewalks with EMC gang signs, as long as you guys wash it off I’m cool with it and will even have a laugh 🙂

    On the other hand, YouTube videos are only seen if you go click on them, not when you’re walking down the street 🙂

    Anyway, enough navel gazing…


  47. @Nick – Here you go – a direct answer, but I’ll give it only the basis of asking to read on which discusses why the question (and the whole post/thread) is wrong.

    Watch this with one of your questions (the 2TB cache one):

    A: Yes, the maximum in a VNX platform is a 2TB SSD-based cache (on top of the main system cache). That cache can be hot-added with no system disruption in a few minutes. It is shared across storage pool blades. It is a read/write cache.

    Now, I’ll point out why (IMO), the whole back n’ forth is flawed (fundamentally – so it’s not about one point or another, but the whole whackload). BTW – I know EMC does this too – so not claiming to be “holier than thou”.

    Followup Q to ask NetApp: Why is Flash Cache read only, versus the read/write model of VNX, and why isn’t it shared across storage processors and able to withstand a reboot without rewarming? If there is content being supported by one storage processor, but the host is accesing via the other via ALUA, how exactly does that work? What is the process to add Flash Cache after I have bought and am using my system if I want to leverage the ever-falling SSD prices? Is Flash cache supported in ONTAP 8.0.1 in cluster mode? If not (I really don’t know), when will it be? In adding the Flash Cache modules, am I giving up anything, like IO expandability?

    **Now, before getting apoplectic and writing a FURIOUS response** – the whole question/answer is flawed – in the SAME WAY THAT THE ORIGINAL Q TO EMC WAS.

    NetApp is different than VNX, and vice versa. The reasons are multiple. NetApp’s primary IO flow for write IO is the NVRAM to the naturally more sequential IO pattern that comes from a log-structured filesystem approach (not trying to get into a fight about the details, rather paint the high level picture). This means that write cache comparisons are not able to be direct comparisons – period. Likewise, you have as a core design principle (for more than a decade) the idea of a “shared nothing” (I’d better characterize “shared little” as NVRAM/IB backend are shared) cluster model. That means that resources used by one storage processor are used by that storage processor. That has good and bad. It means you can do somethings that others can’t. It also means that Flash Cache, as PCIe card will inherenetly be used by each brain independently which again, has good and bad. The choice of using a PCIe card vs. a disk has advantages/disadvantages. Yours is harder to add/remove (and consumes PCI slots), ours adds a little latency (though the latency added is noise compared with the latency of the Flash itself). MORAL – we’re different. And every difference has a “pro and con” which is “obscured” not “enlightened” by a trivial “questions to ask” process.

    I’ll go further – it’s fatally flawed in another way, rooted in the same “we’re different” point. We have 2TB of FAST Cache on a VNX7500 – but consider – there is no limit to the amount of SSDs, and we do an automated tier to SSD. I don’t know what the caveats and limits are with NetApp and SSD in the rest of the system. I do know that you don’t have a peer to FAST (and I’m sure would claim you don’t need it). With EMC, when content is aready on SSD, it doesn’t get promoted (and that applies in both directions). FAST Cache is viewed as system memory, and data sitting in an FAST pool is viewed as a storage block, but their overall latency is similar. So, I could “claim” that we support SSD FAR more than NetApp does, and ask “do you support 200TB of SSD in a system? We do”. In the past, RAMSAN was used as an answer on that, but that has in of itself ups and downs, not worth bringing up.

    There’s a second, and very different fundamental flaw in the whole process. The amount of FAST Cache actually isn’t limited in any fundamental way. We’ve determined that 2TB is a good max based on the metadata that is used by and consumes system memory. We could change it at any time (in a small software update, or by upping the limits of memory in the system).

    That means that above and beyond the fact that the comparison is flawed fundamentally, even the factoid itself (which we’ve already discussed is irrelevant) at the root of of the question is likely to be wrong at an arbitrary future point – but the blog post is there for all eternity, basically ensuring that your readership is running around with a loaded gun, not sure where to aim it. Again, because I’ve been blabbing for a bit, it’s worth restating – EMC does this too – not claiming “holier than thou”.

    Perhaps I’m tilting at windmills, and it’s impossible to lose “competitive tech marketing” in our business, particularly from the smaller folks at the larger folks (which then engenders retailation). I’ve had this discussion with Vaughn, many, many times, and lose it ever time in my mind he “falls of the wagon”.

    Lastly – anytime either of us (or any vendor) is wrong on any ONE in front of a customer, everything we say afterwards is USELESS, as we lose all credibility.

    This is why I encourage our competitive teams to spend less time on “checkbox lists” and “questions to ask lists” (though these are important to know and TRY to keep current – more to track gaps in ourselves than those of the other guy) and more on “what you need to do to be more competitive” (which is something the field can ALWAYS use).

    I encourage them to view themselves not as “negative competitive teams”, but “positive EMC competitiveness” teams (in other words how do we make sure we put our best foot forward with the customer, and in that process, “innoculate” the campaign from the silly competitive lists that always come up).

    So – to net it out, personally I think the whole argument and “questions to ask” lists are fruitless. We’re DIFFERENT – and it’s about the customer, not about us. These “list of competitive questions” are no more than “rock throwing exercises” that our competitive folks do, and sometimes, we ourselves as technical folks – thinking there’s some fatal flaw in the other guy (and each of our continued market success, suggest neither of us are that bad) do.

    I’m not going to post anymore (probably a good thing for the poor reader – my comments are long and boring), but just wanted to answer one of your question very directly, then use that to point out how the whole thing is wonky.

    And – in the positive spirit I’m trying to engender, here’s a virtual hug for you 🙂 (picture of me in Barcelona giving them out on the street with my EMC vSpecalist peeps):

    The last word is yours amigo…

    1. That’s some reply! I also saw the post in your site, sorry if this whole thing gave you heartburn.

      Sure, this post won’t be accurate forever as both our architectures evolve. Just yesterday I was asked by a customer if the Dell/EMC competitive paper “benchmarking” NetApp (back from 2006) is accurate. I had to spend 2 minutes writing an email responding to that. Not a big deal.

      Here’s what I’ll say: I’ve actually seen you present and, interestingly, you had very similar questions interspersed throughout the conversation. Sometimes extremely specific, depending on the subject. Always prefaced by “I’m not an expert but…” – which is kinda funny after reading your response.

      I also noticed that the typical EMC answer to advanced features other vendors have is “minimize and divert”.

      “Snapshots? Sure, we have them too. Look, I just took one.”

      “RAID-DP? That’s just like RAID6, everyone has that.”


      Of course, that’s almost like saying “the competitor is selling you a Ferrari with ceramic brakes? Sure, our car has brakes too, that’s not a differentiator. See, I just applied our brake. And not only do we have brakes in our cars, we even have an emergency brake!!! And a steering wheel for those difficult moments when you actually need to change direction”.

      The one that always irks me:

      “Can EMC deduplicate VMs?” is the simple question.

      The answer – depending on the engineer/rep – either is “sure” (which is confusing and inaccurate as we all should know) or goes to the slightly more accurate but still evasive:

      “You don’t need to dedupe VMs, just make sure you put customer data inside a NAS share instead of in the VM and we’ll dedupe that”.

      EMC’s stance is that this should be enough to deal with the bulk of the data.

      Guess what – not only can you not always do this (indeed, it’s mostly applicable to VDI workloads), it’s actually a lot more complex than just turning dedupe on for the whole datastore.

      Most of my customers have VMs that have nothing to do with VDI. Just a bunch of normal Windows Server VMs that do a variety of things (with sometimes significant amounts of data inside them – they’re not just a bare OS with a couple of services). So I’m not talking about some bizarre edge case here.

      With NetApp gear they simply turn dedupe on and get pretty good savings. This is just impossible (currently) with EMC gear. It’s OK, I’m sure you’re working on implementing the feature, but the current “workaround” doesn’t cover all use cases.

      Not that the suggestion of using NAS for VDI data is bad at all, it makes complete sense, absolutely you should use shares and some thinapp goodness, I’m just saying it’s not applicable for all cases.

      The problem for customers is that they’re now told by everyone that everyone has the same features.

      Everyone can do snaps, and RAID6, and replicate, a few can compress and do some kind of dedupe, flash is now in play with most vendors, etc. And the sad truth is that most customers will check off the “snapshot” box without thinking more about it.

      Short of doing a full-blown bake-off, it’s impossible for any customer to choose, since every vendor pretty much claims to be just like NetApp, and adds a bit of FUD and their own differentiators.

      What this is doing to our industry is that it renders all products into commodities and the cheapest/closest vendor wins (after all, “they’re all the same, I might as well hook my buddy up since he takes me golfing”).

      Since pretty much everyone is doing this kind of selling now, relationships are more important than ever.

      But I do absolutely agree that this should be about solving the customer’s problems, not a tech pissing contest. Sometimes NetApp will solve the problems better, sometimes EMC, sometimes it’s a wash – which partly explains the continued success of both companies versus everyone else (the replication marketshare stats, for example, are comical – NetApp and EMC have about 30% of the market each, and everyone else is fighting for the rest).

      I just wish some customers knew what they were missing. Ignorance is, indeed, bliss. And the grass is sometimes greener on the other side.


  48. Full disclosure – 13 years with EMC. Many as a software architect. Lead on Powerpath and ATF.

    Now engineer and NetApp enthusiast.

    NetApp has – probably – less than 10-20% of the code base of an EMC box. Maybe less.

    The EMC code started as a raid driver built for an old DG system, ported onto Windows, contains a tightly-coupled pair of controllers to support HA mashed in, glued with a NAS gateway, and then additional layers of abstraction added. Its not unified. And still uses copy-on-write snapshots.

    A NetApp is not managed with a single interface. But whoever mentioned “Filerview” is out of touch as its never used with 8.x ontap. But yes, I use some command line, some Systems Manager, and DFM/Operations Manager. Systems Manager really isnt stricly necessary – Many hardcore NetApp folk go straight to the shell. Yes, there are a lot of interfaces, but not because you need them all, it is because you simply have options.

    The VNX is not an incapable system. In fact its pretty good. Evidence shows you can run a a large IT operation using these boxes. I know many engineers at EMC who have poured their hearts into that box, and they are a great bunch of people who are highly talented. The VNX is be fast and reliable, and is truly an enterprise system. But it is a heavy design.

    The NetApp is so light and flexible. And the controllers and cache modules are very loosely coupled, making head swaps and part replacement easier. Everything builds on the pointer-based snapshots, allowing for a high level of code reuse within the architecture. And the fixed block dedup??? I love it.. so easy, and really such a simple modification to ONTAP. 24-disk raid groups? No problem! Amazing how little emphasis needs to be placed on tuning RAID groups with a pooled storage system.

    In summary, the VNX is good. Its just that the NetApp is pure unified storage, and the VNX really isnt. The VNX is good, but in my opinion, the NetApp is better.

    I suppose you would have to be a software engineer, in particular a systems level kernel developer and/or a driver engineer, to really appreciate it.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.