So just the other day, CRN generated this list of 8 storage virtualization products. But that list is pretty plain vanilla, and while there's some stuff missing, I have to wonder if everything actually even qualifies as well. [By the way, Steve Norall has an excellent article on storage virtualization on InfoStor]. Really this is indicative of what I think is a bigger struggle that we're having as an industry in classifying solutions.
I alluded to some of this yesterday, as a prelude to a discussion about iSCSI, but maybe not real clearly. IP SAN's were my example yesterday, and I mentioned that in my opinion, we don't have IP SAN's any longer. The things we lump into this category are actually just SAN's, or storage virtualization products, or arrays. I think we're really just fooling ourselves with IP SAN's. Sure, you might have a full fledged IP SAN, but does the IP nature of it really exhibit any fundamental difference from a FC SAN? Especially if you integrate some IP functionality into that FC SAN? We ended up accidentally creating a pseudo-category, which blurs our ability to separate out functions and really assess the appropriateness of the solution. This seems to me to make it hard for the vendor and the customer. Instead of playing against the field of virtualization solutions, or arrays, these vendors get lumped in a separate category, frequently with the perception that they must play as 100% of the environment rather than partial integration, and as a result, they sometimes fall off the radar.
Because we're struggling with so much innovation and the aggregation of so much functionality into single solution packages, we're losing our ability to group sets of functionality together and figure out what really plays in each space. Hu Yoshida over at HDS, if I can take some liberties with a recent post of his, indirectly alludes to some of this in aggregating services into front-end and back-end. The problem we're running into, is we don't have categories anymore. This makes it difficult for the practitioner to build and integrate solution stacks if you can't capture all of the functionality you need in one solution stack from one vendor.
But I think that's why we're missing some vendors, we can't separate out what's what anymore. To cut to the chase, here are a few that I think are missing or mixed up:
File Services Virtualization - because of collapsing functionality, you can now fully virtualize and manage storage in "only" the file layer, while pretty much letting the block layer underneath be managed by file servers. This requires 100% investment to fully reap the benefits, but with some big players in this space, you can't leave out of consideration the NetApps, Acopia's, and NeoPath type stuff.
Incipient - which I still think is the market leading example of network based virtualization, which should mark the way of the future (you can see some previous posts of mine on this topic for why I think this is an important trend).
HP - HP made the list, but still lacks anything outside of controller virtualization. I still think, with no real evidence, that big things are to come in the future here, but there's nothing there yet. Look to couple HP arrays with a file services virtualization front end, or some type of clustered file system to really enable virtualization.
Sun - I have to question Sun as well as a leader, especially on a list of 8. I think they have some great capabilities in the 6920 product, but I'm looking for a bigger vision from Sun, something akin to virtualization controllers, or a fabric based solution like Incipient. From my angle, the capabilities in ZFS almost overpower anything else, but still doesn't count as virtualization, which I would expect to support more than one platform (maybe when my Mac has ZFS, I'll change my mind).
Finally, "IP SAN" Vendors - So Compellent made CRN's list, which is great - they're an array vendor which can interop with some other arrays to deliver the same functionality across heterogeneous storage. But back to my first comments on IP SAN's, I think we unfairly leave these mis-labeled IP SAN vendors out of the picture, along with a bunch of array vendors as well. LeftHand, EqualLogic, and Intransa all have some great, modular arrays which virtualize storage across lots of storage. That functionality might be limited in application to only limited storage, but it really marks some thought leadership from these folks on what virtualization can deliver, and it's still storage virtualization, just not heterogeneous (although LeftHand runs on HP these days, and wouldn't it be interesting to see that grow into bigger backend storage?). If you're shopping for a bunch of all new storage, there's some compelling value to be had from these folks.
So back to this struggle with modularity...
Even though it's painfully against my nature, I have to wonder if in the current market where this is more the practice than the exception, whether the best choice for most practitioners is to pick a single vendor to do your shopping from and give up on unavailable features which might be key enablers to your enterprise. If so, that's regrettable to me because I believe we stifle innovation, and likely get less than the best capabilities for our businesses (measured of course against the business value of those capabilities).
From that angle, that's why in a previous job I latched onto Incipient when I first got my hands on their product. When I look for solution components, I like to be able to integrate intact, replaceable services in my environment. Incipient captures a full virtualization layer into one layer of your solution stack. It's the fundamentals of virtualization - host independent data copying, volume striping, resizing, migration, and other tools - but it's only very loosely integrated into your environment. Sure, you can do some pretty advanced things with it, network volume striping for example, but if you want to pull it out of your environment, migrate the volume to another volume, and presto, it's gone. Today, that's a unique level of versatility and what I'm calling modularity (freedom from hooks into other parts of your environment).
I'll still at least hope more vendors will think about their solution stacks in this way. What if the practitioner just needs this one capability from your solution? What if you go out of business? It's the 5 to 10 year vision that matters to me, not just paying the piper today.
...Writings here are the loosely constructed opinions of one man, and do not necessarily reflect the views of the Taneja Group.
Read or comment on the original, here.