Home | Storage | Performance Tier or Capacity Tier? – No It’s the Working Set
Performance Tier or Capacity Tier? – No It’s the Working Set

Performance Tier or Capacity Tier? – No It’s the Working Set

8 Flares Twitter 0 Facebook 0 Google+ 8 StumbleUpon 0 Buffer 0 LinkedIn 0 8 Flares ×

There have been a couple of articles on The Register over the last couple of days talking about EMC CTO John Roese and his use of the terms “Capacity Tier” for VMAX storage and “Performance Tier” for their flash solutions.  Sadly this shows EMCs CTO has an outdated view of the industry, is perpetuating old-school views of storage to meet marketing purposes and isn’t doing justice to even EMC’s product portfolio.  Here’s why.

The Working Set

Anyone working closely with any operating system knows you have a working set of data on which most activity is happening.  Where possible, you want all this data in memory, as bus-connected memory is very fast.  However for resiliency, data occasionally has to be written to a permanent storage medium and data also changes from hot (active) to cold (inactive) over time, meaning the working set of active data changes.

As we evolve storage, the aim should be to have all of the active working set in memory or on flash, as close to the processor as possible and running as fast as possible, while ensuring data availability in case of things like power loss.  Thinking of “performance” and “capacity” tiers isn’t really relevant as everything over time will be on the capacity tier except for that in memory.  We’re moving to that today with PCIe SSD and a recent announcement from Diablo Technologies, which I’ll cover in a separate post.

The Architect’s View

“Capacity” and “Performance” are traditional ways of describing storage infrastructure.  We’re already moving away from this.  EMCs CTO is stuck between the technology and his marketing department, looking to perpetuate those high margin VMAX licences.  EMC’s strategy today seems to be scatter-gun based – keep the old VMAX and VNX workhorses with their great margins, while trying to build business with Flash, PCIe & technology like ScaleIO.  EMC could do with showing us a strategy, not a marketing portfolio.

Related Links

 

Comments are always welcome; please indicate if you work for a vendor as it’s only fair.  If you have any related links of interest, please feel free to add them as a comment for consideration.

Subscribe to the newsletter! – simply follow this link and enter your basic details (email addresses not shared with any other site).

Copyright (c) 2013 – Brookend Ltd, first published on http://architecting.it, do not reproduce without permission.

About Chris M Evans

  • Lardinio

    Performance vs capacity it seems is the simplified architectural view. Is it wrong to describe it this way? No, not if you are talking about storage features. I suppose your argument (whilst actually agreeing in some way with what John said) is that we describe our requirements based on the state of the data (hot/cold, active/inactive) rather than the characteristics of the underlying storage. And that’s a fair point. There is no doubt we are going to see a real shift toward in-memory/flash computing. The recent glut of announcements – Diablo, Smart, Crossbar – demonstrates an appetite for investment and change. These are interesting times indeed.

    I also agree with your statement around EMC’s lack of messaging cohesion. The problem is (which I am sure you are acutely aware) is of legacy installed base/high margin service cannibalisation. EMC, in my opinion, have done a decent job of getting the innovation message out there with recent announcements and acquisitions (ViPR, ScaleIO etc), and certainly better than NetApp, but clearly they are concerned about leaving their customer base hanging out to dry, so ViPR becomes a convenient way to sweating the Symmetrix asset a little more.

    • http://thestoragearchitect.com/ Chris M Evans

      Lardinio

      All fair points. I’m in the process of writing up my review of Diablo. We are seeing innovation and evolution in storage. In some respects, to quote the work of Andy Grove, EMC needs to ensure they’re not at an inflexion point and subsequently get left behind. Always a difficult balancing act.

      Chris

  • Pingback: penguinpunk.net » EMC Forum 2013 – Sydney – Part 1

  • Pingback: Design considerations for the host local FVP architecture - frankdenneman.nl

  • https://communities.netapp.com/community/netapp-blogs/about-data-storage Larry Freeman

    Nice write up Chris, I agree tiers are old school and ”working set’ is a good descriptor for the active data that should always be accessed from flash. Now we just need a better way to describe all the other data on HDD that’s not in the working set (the unemployed set?) BTW it would be useful to have a tuning tool that would measure the percentage of data accessed from flash vs HDD. Any data accessed from HDD is a missed oppurtunity to increase throughput. I would think a well-tuned hybrid array today would see about an 80% flash hit rate, will get better over time as flash caching algorithms improve.

    • http://thestoragearchitect.com/ Chris M Evans

      Larry, what figures do NTAP have from their use of Flash Cache? Admittedly this is read-only, but there must be some sizing numbers about the most appropriate cache size that could serve as a guideline. Do NetApp publish that?

      Regards
      Chris

  • Matthew Yeager

    Open Disclaimer: I’m a proud EMCer, fortunate to have been asked to be a Founder Member of EMC Elect, but have been on the ‘outside’ of EMC as a customer, partner, blogger, and various technology positions for many more years than I’ve been inside EMC and have known and respected Chris for a number of years.

    All valid criticisms and, whilst I might disagree with some of your conclusions, I couldn’t help but return … several times … to a critical sentence, ‘EMC could do with showing us a strategy, not a marketing portfolio.’

    Hmmm. Okay, pithy and grabs the attention. I get that.

    But one wonders what definition of ‘us’ you’re driving to as, given you are a member of EMC Elect surely you could ask for a strategic review with EMC executives?

    As the market matures, bifurcates, twists and turns it can be challenging getting certain core messages out there, hence EMC Elect was (as one basic tenet) brought into life not as a cheerleader for all things Pantone 294CV but, rather, to elucidate and educate beyond the marketing and, yes, criticise the very hand that swags us when needs arise.

    I was (and remain) in awe of Joe Tucci’s EMC World 2013 keynote when he threw down the gauntlet (in my opinion) of customer centricity within EMC stating ‘We will provide choice to our customers. And you know, a bunch of EMCers are going to boo me … but I’m okay with that.’ All whilst HDS, NTAP, IBM, and HP logos were listed in full view behind him on a screen the size of an American football pitch.

    My point is not that you or anyone else should be drinking the Kool-Aid as EMC Elect (I certainly haven’t), but provide the leadership I agree wholeheartedly our industry requires to continue moving forward.

    • http://thestoragearchitect.com/ Chris M Evans

      Matthew, thanks for the comment. So far I haven’t really understood where EMC Elect is going. I expected that there would be regular newsletters and other content and that the interaction would have been more of a “push” than “pull” relationship. It would be worth having a discussion about that separately. Perhaps there is also an opportunity here to discuss strategy with EMC and I would welcome it.

      Putting that all aside, I have to admit I struggle to see where EMC is headed. The company now seems more like a product portfolio holder than a company with an overall strategy. Now, that may not be a bad thing; EMC have bought into markets as they have become popular – VMware being the ultimate correct purchase. However recent product releases haven’t been groundbreaking but market following – Syncplicity (Dropbox, Box etc clone), XtremIO (SSD array clone), XtremSF/SW (PCIe SSD clones). Perhaps the only one that stands out as different is ScaleIO and I’m not sure where that fits yet. So, claiming to be “leading” the market is somewhat disingenuous. That gets back to my original comment “show us a strategy, not a marketing portfolio”, which I still think is justified.

      Let’s talk offline, preferably over a quality coffee or beer and keep the conversation going, as I think that’s the most important point you’ve raised here.

      Cheers
      Chris

  • Pingback: Storage Innovation – Diablo Technologies | Architecting IT

8 Flares Twitter 0 Facebook 0 Google+ 8 StumbleUpon 0 Buffer 0 LinkedIn 0 8 Flares ×