Posted by: Matt Stansberry
Blade servers, converged infrastructure, DataCenter
I recently talked to Jonathan Eunice of Illuminata for a case study I wrote on the HP BladeSystem Matrix. We talked about the benefits and challenges associated with converged infrastructure platforms, aka Infrastructure-by-the pound.
This is an excerpt of the conversation that didn’t make the article:
Why wouldn’t IT departments build exactly what they need out of commodity components instead of getting locked into a system like this?
Jonathan Eunice: For most companies, if they did the world’s best job of selecting, integrating, and configuring the components, their top or bottom line advantage would be? Zippo. They just don’t do enough IT to have a little better assembly of the components than their competition make a market difference. So they’re much better off spending their IT selection, construction, and integration time/effort/money on things that can deliver business benefit–selecting apps, building business logic, etc.
It’s much the same reason that most companies that operate fleets of cars or aircraft don’t buy components or kits to assemble themselves. The real business value isn’t in the assembly, for them, so better to contract with a vendor to deliver the infrastructure ready-made, ready-to-use.
So completeness, integration, and having someone else responsible are good. And now the trade-off: It makes you more beholding to your vendor(s). Lock-in is an inherent trade-off of ready-made. What’s that saying? Pick your poison?
I talked to a BladeSystem Matrix customer and he mentioned the self-service provisioning is making VM sprawl a problem. Can you see this becoming a common issue? How do you keep it reined in?
Eunice: IT sprawl is a fact of life. Enterprises’ employees, divisions, and departments want a lot of IT to be accomplished. Tons and tons. VMs make it easier — no waiting weeks or months for new physical servers to be approved, delivered, and installed. Just turn a VM on! Presto! Self-service provisioning is a recipe for even more IT to be deployed. But if you don’t let business functions do self-service, then IT is the bottleneck to the business getting the IT it wants. There are really important trade-offs either way. But they are business and governance trade-offs, more than they are Matrix or infrastructure-by-the-pound trade-offs. Because virtualization and self-provisioning are the technical enablers, they get blamed for what is, in fact, a business policy and governance issue.
VM (or IT app) sprawl is a much smaller and more manageable problem within enterprises who’ve tackled things like governance, policies, asset inventory, and change management up front. If deployments get ahead of such procedures for handling IT at scale, sprawl is the typical result. Sprawl = scale without appropriate management and control.
Are the Cisco and HP products comparable enough to lump them together as “Converged Infrastructure”? Are IBM and Dell scrambling to keep up with something in this category?
Eunice: I think of Cisco UCS as a bit more of a building block, akin to HP BladeSystem or IBM BladeCenter. HP’s BladeSystem Matrix, IBM’s CloudBurst, or Oracle’s Exadata are a bit more fully integrated, wrapped, and packaged as “plug it in, turn it on, go!” solutions (for different kinds of workloads). But that’s a bit of a taxonomic nit. UCS came last to the blade server game, so the way it’s pitched by Cisco or perceived by customers is a lot more colored by infrastructure-by-the-pound and virtualization-everywhere thinking than the foundation HP or IBM blade products. If customers want UCS compared to Matrix etc, I do that. If they want it compared to basic blade servers, I’m happy to do it that way too.