Monday, July 7, 2008

Virtualization: Panacea or Pandering

After reading a few seemingly unrelated blog postings, I let my defective brain begin coalescing the pieces into a coherent thought. Dangerous stuff for me to do. Of course, it's likely flawed, but it might be interesting. Let's see...

The general focus here is whether "virtualization" is truly the wunderkind of technology that the marketing forces would have us believe. Maybe. Maybe not.

First, was the post I mentioned a few days ago about Microsoft product licensing for idiots. Then was the blog post by Mark Mac Auley on "What isn't being virtualized?" After mixing these two ingredients on low heat, I added a pinch of coffee mess discussion with a coworker who completely discards virtualization as marketing hype. He says "show me the savings" and dares anyone to cross his line of logic. This guy is sharp and starts his caffeine intake before most of us wake up. You have to be prepared for him if you want to do battle over technical stuff (yes, Rob, that would be RR I'm talking about).

The somewhat obvious shifting of costs would imply just that: a shift, rather than a true savings. Things like beefing up the hardware and networking to concentrate more virtualized processing on a given physical host. Other things like purchasing (or reallocating) the same number of Windows and Office CAL's from distributed nodes to virtual nodes. Again, no real clear savings. Moving server services like SharePoint, WSUS, even AD domain controllers, offers no clear licensing savings.

So, then you bring in the benefits of centralized storage (assuming you drank the same Kool-Aid about HP, IBM, NetApp, EMC or whatever) and the portability and failover aspects of virtual machine management (SCVMM or VMotion). We're talking site-to-site or intersite here, because intrasite doesn't offer as much failover due to shared facility infrastructure resources (I'm talking plane crash, building fire, hurricane actually).

Sure, there's some added flexibility and increased availability, but that too comes at a cost. In some cases a huge cost. Intersite bandwidth first of all. Then comes redundant site configuration (server, rack, cooling, power) and of course: redundant host licensing (Vmware, host OS, management tools, support, etc.). Let's not forget training and contingency planning and testing. Pause: If you run VMware and VMotion - when was the last time you actually pulled the power to your main site to see how this works? I suggest doing it on a holiday weekend, without giving any heads-up to the staff. :)

Then comes things like VDI. My coworker just flips out over this subject. Going back to the shifting of hardware, but then turns back to remind us that there's an inherent advantage of having distributed processing: If the network goes down, we can still work locally. And when he picks on WAN availability to the remote VDI host site, I counter with VDI brokers and regional host servers. He counters with added cost for all the distributed servers, infrastructure, access control and square footage (lease rate). The VDI client/server issue can be addressed early on with planning. Then comes "VM sprawl" and additional costs and tools required to maintain centralized control over the entire virtual landscape.

We're not even touching on virtualization of networking or storage here either.

The net result is that at 100,000 feet, it's very difficult to provide a tangible cost savings which is all-encompassing. It's a mix of hardware, software licensing, connectivity, management, storage, infrastructure (electrical, cooling, space leasing, access control) and so forth. It's more of a complicated factoring than most pure-play IT folks would expect. The takeaway from this is to do your homework before you walk into the lion's den (MBA staff meeting) and say things like "it will save us money". If they know anything about business management, they will demand detailed analysis for your business case.

So, when the vendor ties your cute little Sponge Bob bib on and begins pouring the Kool Aid down your face and neck, stop and think about the peripheral aspects here.

Yes, you can add incredible flexibility and redundancy to your environment. But it's certainly not free. There will be obvious implementation costs to deal with. But there will also be recurring administration costs. Be sure to factor in everything that relates to your plans to be sure you don't end up apologizing and begging to keep your job later on.

Don't get me wrong. Virtualization is definitely powerful and useful stuff. I strongly encourage everyone to consider it. The key word is "consider". If you're not at least giving it consideration (translation: testing and evaluation) you are cheating yourself. If you come out of a vendor luncheon/demo/presentation/conference and immediately fire off purchase orders without having spent at least a month or two in rigorous testing, you need to stop and rethink what you're doing. Just like a good power tool, it has it's ideal uses: Virtualization is perfect for some situations and not really good for others.
Post a Comment