10 data center paradoxes -- out of complexity comes simplicity

TalkBack 1 of 4:: Next »

Thread View
Flat View

Out of complexity - comes profit: Consolidation is key to squeezing the most out of your IT dollars. This is the universal "need" that all companies have today. Actually it's been a need for as long as I can remember. But in the past this need has been ignored - because of POLITICS.

In the past, customers and beancounters prevented consolidation. Here's an example of what I mean. Project A gets approved for $10 million and buys infrastructure. Turns out that Project A only uses 25% of the capacity of its servers. Project B is an important "legacy" application that is "starving" on old infrastructure that it runs at 100% much of the time. Since Project B has no budget (remember it's legacy), it cannot buy new infrastructure - but it's used every day and is very important to the corporation. What can be done?

The answer seems simple - consolidate both projects on the new hardware. There is ample "bandwidth" to run both projects on the same server - so why not? AhHA! The department that created Project A thinks it "owns" its servers and says NO you can't do that! The beancounters cannot figure out "utility" pricing and since it cannot charge-back properly - it can't happen. As this "condition" happens more and more, the datacenter starts looking like a "rummage sale" with many brands and models of servers on the floor (because each department was able to buy whatever infrastructure they wanted). A previous manager of mine called this condition "One in a rows".

So the customer and beancounters have created this mess without letting IT "fix" it for them. The simple process of consolidation (moving apps to underutilized servers) becomes impossible because of political constraints.

Virtualization changes this dynamic. The beancounters' job gets easier as they only deal with one vendor (does that mean we can fire some of them?) and can use utility pricing. The customers like it as they ask for something, have the budget, and it happens! The IT department likes the new equipment (and the skillset-enhancing resume-building that goes with virtualization) - but their job gets HARDER. If an admin had 100 servers to handle before, he still has that many (if not more) VM instances. He also has the overhead of managing the VM system (or systems) itself. For Windoze the VM world is great (since Windoze servers cannot run more than one thing at a time), but for *NIX it's just stupid. All of those years that *NIX perfected handling multiple processes is thrown out the window - as you only run one app at a time. This means that load balancing that used to be automatic under *NIX becomes a MANUAL PROCESS under VMs. This burdens the sysadmin, and he reacts as any human would, he ignores it. This means that those VM instances are rarely adjusted and could be running at 10% capacity. Full-circle!

So what you have now is a bunch of EXPENSIVE new servers that look nice and attractive sitting on a datacenter platform. But on the inside, they are just as inefficient (or moreso) than the equipment they replaced. But out-of-sight does mean out-of-mind as a stroll through the datacenter won't show you the problem anymore. So if you don't have a very active Capacity Management team that holds sysadmins feet to the fire - nothing will get done and things will still be inefficient. So is THAT what you paid for? Is that why you want Virtualization?; Posted by: Roger Ramjet Posted on: 06/18/09 (Edited: 06/18/2009 @ 10:51) You are currently: a Guest | Members login | Terms of Use