We can build a server in 17 minutes said the nice man

I was for the second time meeting with a consultant who works in the Wintel space, I was wanting to speak to him about build technologies, standards and his views on automation best practice, an overview if you will to get other views and see what’s going on.

Anyway, it was all very interesting he was mentioning interesting tools like Opsware, Altiris and BladeLogic, what innovations they had made from operating system load through to deployment into production which sounded stunning and it was at this point that it all went wrong and I will explain after the following quote.

“We can build a server in 17 minutes”

Fantastic I replied. That’s brilliant news and shows you with the right procedures, the right technologies and best practices we can as an internal IT team be as efficient or nearing so as an external vendor. 

What do you mean by build I asked?

Installing the operating system? Why what did you think I meant?

So how long does it take from start to finish? I asked.

There was hesitation, dare I say it a bit of awkward silence as if I’d started speaking to him about the Renault 6.

Well it takes seven weeks.

Sorry, I ask?

Seven weeks, you know by the time we put the specifications together, raise the acceptance, sign off and then raise a purchase order, then a week or so to rack, then build the OS in 17 minutes and then hand over for the application load, then four days for the monitoring, backups and support handover to take place.

So what operating systems can you build in that time? I ask

We can build Windows 2003 and Windows 2008 all versions from our template image.

If I want to rebuild our old Windows 2000 server? I ask, You know if the server goes bang, how long will it take to rebuild.

Oh we would need to burn a CD and then get the drivers and build it manually.

Why? Well that’s out of scope, IT and the build team don’t support Windows 2000, it’s on the risk register and is therefore at risk, if the system failed we would deploy a Windows 2003 server and ask them to try that first whilst we fix their broken one.

It’s at this point, my interest in the conversation started to wonder. Don’t for a moment let me diminish the achievements of the IT teams, any streamlining efforts in the build space made are to be commended, any efforts at illustrating the ability to deliver an agile platform to meet the business need is brilliant news. However, in such scenarios we seem to have ‘danced around the edges’, by this I mean I want a holistic offering, an effort that attacks not only the build part, the actual putting the Windows CD in the drive and pressing go, I want to streamline the architecture, the sign off, provisioning/rack and stack through to application load.

We need to re-engineer the server provisioning process to encompass everything from the purchase of the server through to the racking, the configuration and operating system load as well as the application load and go-live.

We need to remove the complexities, data gathering and teams involved in buying a server, at the same time we need to write in the sand and have the relevant people sign off on what is and what is not supported.

The build platform needs to be able to deploy future operating systems, the Windows 2008 and beyond, as well as the in support but not being deployed Windows 2000 servers that might still be humming along but not harming anyone.

The provisioning model needs to be simplified to:

• Standard offering – the automated delivered in weeks or hours solution based on a range of standard enterprise servers with default configurations which we buy in bulk (alongside expected forecasts)  – with this we deliver in weeks or hours as we have them in stock, we’re repeating known processes and there is no ‘buy-in’ or specific complexities in deployment – it’s a server with or without SAN, with network, with lights out and an OS. The process would be:

• Commit to purchase server.
• Rack server
• Apply fibre/ethernet connections
• Initial hardware configuration
• Network boot and OS load
• Handover to app team
• Application load
• Go-live

• Specific offering – the formalized front door provisioning model where everything is manually request:

• Commit to a server
• Agree specifications and CIO acceptance to deviate from standard offering
• Commit to buying server
• Raise purchase order
• Raise install request
• Rack server
• Apply fibre/ethernet connections
• Initial hardware configuration
• Operating system load
• Handover to application team
• Application load
• Go live

Category: News, Servers, Virtualisation