Unified Computing: The Day After the Day After
Now that the dust has settled and everyone has had a chance to digest the Unified Computing System announcements a bit, I see a couple of salient points–well, I see more, but only have time to blog about a couple of them right now.First and foremost, we have certainly move the bar for everyone. As pointed out in a SeekingAlpha post by Larry Dignan:
Other data center rivals will have to start answering questions about integration between functions, architecture and the potential savings. If anything just the buzz associated with Cisco’s move can dictate the conversation.
In the long run, I think this will drive the industry forward–as Doug pointed out recently, “we are certain over time we will not be alone in this market.” In the near term, however, I see a lot of talking points on why this really isn’t any different or really all that important.The second point is that the “S” in “UCS” stands for system, not server. The UCS is an integrated platform, built and designed to be operated as system, and this is where we believe the value is created for the customer. The ability to manage systematically and holistically is key to virtualization adoption and TCO reduction. Let’s look at another integrated system you are all familiar with: an automobile. If you push the accelerator, you go faster–its a simple model, you focus on what you want to accomplish not how to accomplish it. Now imagine a scenario where, to go faster, you had to turn one knob to increase output from the fuel pump, press a lever to increase air intake, turn another knob to adjust fuel/air mix, turn another knob to adjust spark timing and finally watch the tachometer and shift into the correct gear–oh, yeah, you also need to watch where you are going, steer, and be ready to jump on the brakes. Its a lot of work to accomplish a basic task, and as a result, you are less likely to change speed. Well, its precisely the situation we have in the data center today–operational complexity and manual intervention are an inhibitor to boarder virtualization and the cost an functional benefits it can deliver.Which brings me to the third point (OK, I lied) about these wild claims of vendor lock-in. This makes no sense to me–if you buy an HP BladeSystem, I assume you have no expectation to plugging a Dell blade into it–this is intuitive, so why is the Cisco UCS being held to a different standard? It is an integrated system. If you don’t want to buy the system, you are more than welcome to buy the Nexus 5000, Nexus 2000 and (eventually) Nexus 1000V separately and do your own integration. The related assertion is that the Cisco UCS will not work in a heterogenous environment. This is a completely misguided notion. There are a number of management models for the Cisco UCS to allow it integrate into both greenfield and existing environments–more on that in an upcoming post.