Blade Watch

We often hear about the need to adopt virtualization or server consolidation (or both) to be more efficient with our data center space and power, to reduce the number of servers we have, to bring us a more dynamic infrastructure, abstracting the application from the underlying hardware. Before we do this though, we need to verify our inventory not just in terms of the servers we have, linked with the applications, but with the applications we have with the servers, so that we can clean and verify the data we have, to ensure that however we virtualize, that we understand all the elements of the ‘production’ or application chain, the feeds, the storage and server as well as the network.

Let’s step back then for a second and use the example that I am virtualizing a data center comprised of 1500 production, development and uat servers.

• Obtain a list of servers and applications compare with a list of applications and their servers – you would be surprised what didn’t add up – oh that’s not a web server for FX, it’s shared – I can decommission it then?  No, it’s an FX server, part of our core web platform for staging. This is crucial particularly in business where cross charging is in place, where you pay per server, there is often the situation where we’ll make everything shared to reduce the server count, the support cost, but this can affect your strategy.
• Decide how we virtualize – doing in-house IT, validating the concept, getting technical backing and support can prove invaluable before starting on ‘real life applications’. Do we virtualize everything by hardware platform? By data center location, environment like production or development, or do we carry out a per application virtualization process encompassing all platforms? Which would enable the most volume of virtualized hardware at the least cost?
• Consider the process, ownership and next stages, we might start off with 3 ESX or Hyper-V servers, what happens when they’re getting full, who pays? What’s the lead time? What’s included in a virtual machine? How do we sell it internally? We need the help desk processes, the how do I, what happens when, billing and chargeback maths so it’s affordable, avoidance of the marginal cost, one virtual machine must not cost a new ESX servevr.
• What are the barriers to entry on production? What do we need in place, are their ownership issues in terms of the hypervisor layer, or monitoring that’s needed, what level of shared virtual infrastructure do we have? If we have several shared virtual infrastructures, for example one for Asset Management, one for Risk, another for Back Office, what are the rules of engagement? How do we stop one business line’s lack of investment from impacting other business lines and avoid any perception or delivery issues with the platform?

Related posts:

1. Going Green with IT and applying a tier to the application and infrastructure Greenme Moores Law states that the number of transistors on...
2. Talking about taking virtualization to the next step Techtarget Now that virtualization deployments are creeping out of test...
3. Virtualization projects need focus and buy-in I was interviewing one of the AtoV consultants about a...
4. Quick virtualization guide – for those thinking about it One of my favourite conversations was had yesterday with a...
5. Virtualization without the emotional response A brilliant example of this was brought to my attention...

Related posts brought to you by Yet Another Related Posts Plugin.