Talking Storage with Jerome Wendt, DCIG: Part Two

I sat down with Jerome Wendt, founder and principal at DCIG, while he was in town to attend the Dell Storage Forum.  DCIG is an analyst firm that is heavily focused on data center technologies, with deep roots in the storage industry.  In this conversation, Jerome shares some of his thoughts on the flash frenzy and as well as the interesting things end users are telling him.  The following is the second installment of a two-part conversation with Jerome (Read Part One).

Don: I think I'd be remiss if I didn't ask you what your take on the flash frenzy was. It’s good to hear that you are looking at an SSD Buyer's Guide. What's your overall take after speaking to a number of companies in the space and what you're seeing in the industry as a whole?

Jerome: Right now it's really hard to even quantify where to put flash. Where's the right place to put it? You've got the Fusion-io camp that says it's in the server (as does EMC with its VFcache.) Then you have Violin Memory that wants to put it in the network where it is a shared cache sitting out there. Then you have an appliance like Nimbus Data or Whiptail. And of course you've got EMC just buying Extreme IO, which is not even technically producing a product yet.

So I think a lot still has to play out in terms of what's going to happen with flash and SSDs. I just finished a two interview series with Fusion-io; one with Rick White and another with Brent Compton, and where each sees flash.

Clearly Fusion-io has a very different perspective on the industry as a whole in terms of how it wants to treat flash as memory.  It views it as a whole new tier of memory and you treat flash as the way it's initially intended: as a form of memory.

But, I think in the near term there is still a huge play for SSD appliances.  I just spoke to StorSimple recently.  In talking to Mark Farley who is now there, he said that flash gets very interesting when using it as a cloud gateway. When you present flash to your storage system, you get this really high performance.

But in the case of StorSimple it deduplicates the data, then moves the blocks that are not being accessed out to the cloud. You can have this really high performance fast tier locally, and this infinite capacity, quote unquote, out in the cloud. And, people do not really have to manage storage in the same way that they conceptually had to manage it in the past.

I think a lot has to be sorted out yet in flash with all three of these approaches likely to be used over time and it is premature to say which one is going to be the big winner but it is going to be fun to watch nonetheless.

Don: I know you spend a fair amount of time talking to end users as well. What's your take and what are you hearing come up? Is it still the same age old problems of where to put the data, and on what tier? Or, are we still looking at something more basic around how to handle the exponential growth of the data?

Jerome: I hear some of that. One conversation I had when I was with a user at Interop is that one of the problems it is running into is in the realm of unified storage.

What is happening in the unified storage space is that companies essentially provide a unified management interface, for lack of a better way to put it. They put their block and their NAS under one management interface. But really it's two systems underneath the covers. Initially I thought, “Is that really even a unified system?”

In talking to this user at Interop, he was saying that one of the problems he's running into with a unified system that has both block and file on the same machine is that performance really begins to degrade over time.  When you put both block and file on the same system, that block data gets very defragmented in these unified systems. So now I'm beginning to swing more over to the other camp where these federated systems that have block on one box -- that handle the block traffic on one server within the storage system, and NAS and NAS traffic on the other one, is maybe the right way to go going forward.

People seem to just care about the single interface with unified, and they do not care so much that that the processing occurs on two different boxes. So that's one thing I'm hearing.

In terms of the big data side of the conversation, unless you get in the really high-end organizations, the conversation is more around bigger data at this point where most organizations are starting to consolidate more.

Virtualization is certainly driving some of that. But as you start protecting more mobile devices, I do not think the impact of that has really fully been seen or felt yet. Companies want to store data centrally, but access it with any device. That's certainly going to drive some changes in terms of how people store their data and manage it.

Don: I appreciate your time. I know it's a very busy timeframe for you and a number of people this time of year. Certainly the number of announcements seems to grow exponentially as well as the data in end users' environments starting in March through the beginning of the summer timeframe. So I look forward to continuing to read what you're putting together for end users, both on your blog, which end users can find at dcig.com, and then also in terms of the Buyer's Guides and the information that you're putting out that way.

Jerome: Thanks for taking a little time to talk with me.