More Virtual Thoughts On Storage

Saw a pretty good piece on www.searchstorage.com where Beth Pariseau interviewed Jay Kidd at NetApp about file and storage virtualization.  (BTW, good article, Beth!)

Jay offered his views on a number of topics, including aspects of storage virtualization.

Of course, Jay's views were colored a bit by the NetApp perspective.

Not that it would ever happen to me .. ;-)

But I thought there was a strategic side to the discussion that was worthy of exposing.  I think it shows a contrast in views between two technology companies.

The first thought -- file virtualization

The interview starts with Jay being just a tad disparaging on file virtualization technology. 

Maybe it hasn't been good for NetApp, but it's been pretty good for us here at EMC.

Since we've acquired RainFinity, I've gained a real appreciation for just how strategic file virtualization is, and how much more so it will be in the future.

He's right -- it's a great consolidation/migration tool, and having a single name space is a boon to NAS adminstrators.

But I think it goes a bit farther. 

First, all the explosive growth we're seeing in information ain't happening in databases, it's happening in files.  I think of it as the next frontier in information management -- what do we do with petabytes of files that can save us money, make us money, or keep us out of trouble? 

There's a growing problem on the horizon, and file virtualization is an important part of the puzzle.

Second, file virtualization goes in pretty easily as compared to storage virtualization or even server virtualization.  With the right solution and use case, there aren't issues with performance, or migration, or management, or compatibility. 

The stuff just seems to go in pretty well, and starts doing the job.  Not a lot of drama.

But I think Jay (and perhaps others) miss a key point -- file virtualization shows every sign of evolving into next NAS platform.

Where would you want to put file replication?  The file virtualization device.

Where would you want to put aspects of file classification and tiering?  The file virtualization device.

Where would you want to put newer file security functions?  Probably the same answer.

And the list of cool functionality that really makes sense to put at the file virtualization layer (and not the client nor the NAS storage) gets pretty long, when you start to look at it. 

Next-gen NAS.

Now, if I was NetApp, that sort of technology disruption be a pretty hard pill to swallow. 

But don't sacred cows make the best-tasting hamburgers?

Put more broadly, we see more storage-related (and information management related) functionality moving to the network. 

That's where it belongs.

True for file virtualization, true for storage virtualization.  Or whatever these technologies end up evolving to be in the next few years.  For EMC -- make no mistake -- network-based virtualization is a strategic play. 

Not some widget.

The second thought -- storage virtualization

Jay positions NetApp's vFiler as competing with HDS's Tagma.  Well, yes and no.

Yes -- it's yet another storage controller that's been pressed into new datasheet duty as a storage virtualization device.

No -- architecturally the Tagma and the vFiler are fundamentally different beasts. 

Hint: the Tagma is a whole lot bigger.

He does say something very insightful which I do agree with. 

He says that if you have a virtualizer in front of six arrays, it'd better be as powerful and as good as the six arrays behind it.

I (and EMC) agree. 

Which is why we believe intelligent SAN architectures are the way to go for storage (block) virtualization.  Many folks would have no philisophical problem putting six storage arrays behind a pair of honkin' SAN directors, right? 

Hence one of the rationales for Invista's architecture instead of Tagma, SVC, FalconStor, and ... I guess ... vFiler.

The third thought -- server virtualization

He's right.  The topic is hot. 

Can't wait to see how the VMware IPO plays out, just to see how hot it really is ;-)

But I think the consolidation discussion has moved on to a dynamic resource discussion.  You get one wave of benefit just by using smaller, consolidated containers for your workload.  And you get yet another whole wave of benefit by using those server resources as a dynamic pool, hence VMware's DRS.

The fourth thought -- indexes and directories

Jay says that there's a new area of indexes and directories.  Metadata creation and management, by another name.

Couldn't agree more.  Matter of fact, I think I've written on the topic a few times.

Good thing we've got Documentum, EmailXtender, Infoscape et. al.  Going further, if you're going to be an information infrastructure company, you're going to need tools to classify, organize and process vast amounts of information.

The fifth thought -- who is NetApp?

He gave an honest answer -- they're figuring it out.  I had a sympathetic moment for Jay.

Been there, done that, not a fun place to be -- for either your employees or your customers.

I've found that companies perform best when they have a clear strategic vision of what they are, and what they're not.  For NetApp, if that ends up being "secondary storage", great.  Just be clear.

I think EMC has evolved to that point where we can claim a consistent and coherent strategic vision.  Certainly feels like it to me.  I would offer that EMC might do a better job communicating that vision, but it's starting to happen.

And -- for me -- vision is everything.

Comments

[Chuck Hollis]

I guess I wasn't the only one who found a few bones to pick with Jay's comments.

Check out Storagezilla:
http://storagezilla.typepad.com/storagezilla/2007/04/raining_on_neta.html

[beth]

(BTW, good article, Beth!)

thanks! :-)

[Chris M Evans]

Netapp must be smokin' something to think they can compare their vFiler with TagmaStore. I did a design job (for a company who will remain nameless, but they know who they are) and Netapp *refused* to sell us their own vfiler solution as they couldn't guarantee performance (I was happy I could guarantee the disk performance as it was USP). Perhaps there was a commercial deal in it too - I imagine Netapp make quite a margin on their disks and disk shelves.

[Chuck Hollis]

Hi Chris

All marketing spin aside, the vFiler architecture and the Tagma architecture are two very different beasts. I kind of choked when Jay made the comparison.

Glad to see I wasn't the only one!