EMC finally joining the SPC – plus some advice

Just came to my attention that EMC is finally joining the SPC (Storage Performance Council). As I’ve pointed in their past, their absence from this most standard industry benchmark was puzzling, kudos for rectifying this omission.

I do have some advice for EMC (and all other vendors that have already posted results):

  1. Do show results for your various supported RAID types, not just RAID10 – after all, if most of your customers don’t just deploy RAID10, it makes sense to show RAID5 and RAID6, especially if you want to compare results with NetApp RAID-DP (that’s the protection equivalent of RAID6). This will increase your credibility. The argument that you only show RAID10 since that’s the best performing doesn’t hold water – everyone knows the other RAID types will have different performance levels, providing results for everything will at least enable customers to get an idea of how much of a performance hit they’ll have with the different RAID types with a write-intensive workload.
  2. Do show long-running benchmarks with auto-tiering enabled. After all, if you are claiming your auto-tiering implementation doesn’t hurt and can even improve results, this is your chance to show it.
  3. Enable features like snapshots.
  4. Use your large cache if you have one, especially if you keep advertising that it accelerates writes. It will just solidify your claims.

Welcome to the club!

D

 

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

2 Replies to “EMC finally joining the SPC – plus some advice”

  1. Nice welcome letter Dimitris!

    The required price / performance disclosure will be particularly interesting for customers to finally see.

    EMC will also need to understand the ‘C’ in SPC means consensus amongst competitors. You can’t control this group, and yes other members can still publish results on competitors’ platforms. The SPC is about scientific performance measurement rather than marketing spin.

    Finally, in addition to your list above, I hope EMC will publish with virtual pools enabled so we can see how Thin Provisioning affects performance of a system under load.

  2. Totally agree to Dimitris and really nice writeup

    Also need to consider below points in the tally 🙂

    1. If we really compare that with competition, then we need to simply take result of the FAS6240 and multiply by 4. This is perfectly legal because, this is what EMC has done with their mammoth configuration to achieve such result. And after doing so NetApp will be 53% faster in absolute terms

    2. Need to consider what is the usable TB in mentioned configuration wrt to competition

    3. Also need to consider $/TB and $/IO in mentioned configuration

    Any other comments guys 🙂

    Note: I am not anti-EMC nor biased to NetApp. Just a geek 🙂

Leave a comment for posterity...