 |
We can be a tight-knit group in the storage world.
For years, I've followed the IDC and Gartner market share numbers, and -- other than jockeying for market share points and stack ranking -- it's a familiar vendor landscape. But change might be upon us. Virtualization has completely transformed the server landscape, and the same process has begun in earnest in the network and storage worlds. While the exact path might be unclear, the endpoint isn't: it's a world where network and storage functionality is preferred to be expressed as software stacks, running on virtualized server hardware. Today, VMware wouldn't show up on any list of top storage vendors. But I can make a bold prediction that -- within a few years -- VMware will likely be on that list. To Start WithNot to belabor the obvious, but we live in a world where there are more virtualized workloads than physical ones, and the trend is accelerating. Competitive jockeying aside, VMware remains the clear market leader in multiple aspects. Customers are fortunate in that VMware has historically invested in value-added mechanisms between servers and disparate storage to better bridge the inherent gap: VASA, VAAI, vVOLs and more. There's only one historical precedent for the degree of advanced integration that has emerged between vSphere and external storage, and that's what you'd find on a zOS mainframe.
While VMware has been extremely successful at influencing the storage industry, it hasn't really been a storage vendor per se. Yes, we're all familiar with things like VSA, VDP and Storage vMotion -- all technically storage products -- but they haven't done much to change the landscape. I think that has the potential to change. Here's why ... #1 It's All About Being Close To The Application Just about every storage vendor will agree -- it's all about the application workloads that drive storage differentiation and consumption. Storage vendors spend an enormous amount of effort trying to understand the unique nuances of different applications, and invest heavily to integrate and optimize their products for Oracle, SAP, Exchange and much more.
Nothing is closer to the application in the stack. That somewhat obscure theoretical advantage clearly pays off in the real world: at provisioning time, when monitoring performance and resource consumption, and when diagnosing and resolving inevitable problems. The closer you are to the application, the better. If you're a storage geek like me, the ability to granularly enforce storage and protection policies around arbitrary collections of VM (application) boundaries that can dynamically move and resize -- well, that's been a gleam in my eye for close to ten years now. Goodbye LUNs -- we knew you well. #2 It's All About Converged Workflows And Operations In my early days working with VCE and Vblocks, one of the arguments I didn't win was on the choice of terms. They wanted to create a new category of infrastructure: converged infrastructure. My argument was that the real value would inevitably result from converged workflows and operations: how easy it was to run this stuff in the real world.
So the argument arises: where should this convergence of operational models occur -- server, network, storage -- or perhaps a new uber-manager location? The virtual server proponents have a strong argument: policy is best expressed at provisioning time. While we're having that academic debate, the world is already moving quickly in this direction: many virtual administrators are already responsible for a decent portion of network and storage tasks. I believe we'll see the best converged operational models emerge in vSphere / vCOps / vCD -- and they'll be natural extensions of what's already happening today. #3 The Lure Of Commodity Servers Twenty years ago, storage pros were enamored by the potential of using lower-cost commodity servers to build storage platforms -- and that's still true today. But things have changed: commodity servers ain't so commodity anymore. Blazingly fast, very capacious, big pipes, amazingly dense, built for availability and monitoring, etc. -- things have moved along at an incredible pace.
Yes, hardware costs can be cheaper, but I don't think the appeal ends there. For one thing, there's consistency: use the same building blocks for either compute or storage depending on your needs. One inventory, one vendor, one support model for hardware. Use the same tools you use to monitor server hardware to manage servers running storage software. And so on. Indeed, we've seen a few of the traditional storage vendors move in this direction, albeit in cautious steps. The real challenge -- at least in my mind -- is that all of them have built their business models around selling and supporting purpose-built hardware. And that's a very hard addiction to kick. VMware, by comparison, has built their business by creating hardware-agnostic infrastructure. For VMware, infrastructure is software, not hardware. VMware certainly has challenges going forward, but weaning off from a hardware-based business model isn't one of them. #4 Storage Is Software Today, the vast majority of storage arrays and appliances are built using familiar off-the-shelf server technologies: Intel CPUs, DRAM, 10Gb ethernet and infiniband, and so on. They look like servers with attitudes.
As we've seen with servers, when software becomes virtualized, magical things can happen:
VMworld Is Upon Us
While there certainly will be some neat technology involved ( as well as some obvious customer benefits): none of these will change the storage world as we know it -- at least, in their current form. As a result, I don't think single storage industry analyst will start to consider VMware as a storage vendor in the same breath as more familiar names. That's perfectly understandable. Industry analysts usually require substantial evidence to change their perspectives, and that has yet to come. But if you pay attention to what VMware is starting to announce in the storage world, you'll see that each component is reflective of every principle I've outlined above. It's a compelling case. All that remains is execution :) -------------------- Like this post? Why not subscribe via email? |
| Update your feed preferences | |
|
|
