Blade Watch

I got asked my viewpoints on re-using servers. Now don’t get me wrong operationally it makes sense, but long term, if you think of the IT strategy, of data center capacity, power, support costs, the big picture if you will. Your lack of investment is going to cost you several times around directly or indirectly.

Re-using your old production servers as development, or taking servers for recycling and making them proof of concept for a project or as a platform to test builds/settings has been normal practice for many in the enterprise and SME markets. Ideally as we move to a virtual world we can provision on demand allocating more resources, more virtual machines as they are needed within our the constraints of capacity.

Why do we re-use servers?

• No budget – the task we want to use them for does not have a specified budget, for example to upgrade development from Windows 2000 to WIndows 2008, or for the DL380 G2 to DL380 G6 or virtual machines.
• Unexpected project requirement – application team has an urgent need for a server for something, the application needs a file server, we wont get a new server installed in time or approval to buy a box.
• Urgent business requirement – extra capacity for the corporate web site, we need a temporary machine whilst we order new ones.
• Opportunities presented – we can see where that server could be re-used to solve a problem, testing or monitoring an application or network, providing a new internal IT or customer facing solution which might not justify buying a server.
• The role isn’t demand enough, it “doesn’t need” to be a new server, it’s a file server, a ping monitor or box to store and run monitoring scripts.

Re-using servers presents the following issues:

• Encourages or supports the concept of sitting still which creates costs  – we haven’t validated against Windows 2003, the application doesn’t work on .Net 2.5, all these kind of things create barriers to keeping your infrastructure up to date, keeping in line with support, and creates a support overhead. By that I mean, we get into the situation where we might have applications and business processes dependent on one platform, one operating system and one database, the cost of move being so great, that I have no option but to support DEC servers running an unsupported version of Alpha, with NT4 web servers creating an additional risk to service and support/security overhead.
• Insource/outsource/merger or acquisition, sometimes as part of the outsourcing activity the service provider will ‘buy’ or take ownership of the servers, this can mean from a financial standpoint the assets clock is reset, “it’s not 7 years old, we’ve only taken ownership of it”.
• Standards in place but being over-ruled, as a service provider this can be particularly difficult, “we don’t support DL380 G2’s, yes you do what’s the cost?” The support cost might be proportional to the capital cost of buying a new machine, and the extra overhead – new disks, more memory or network cards that might be needed.
• The older servers are typically:
  • Less energy efficient
  • More of a support overhead – the operating system might be nearing end of life, the hardware will be getting older and more prone to failure, the systems will need managed more and actively engaged.
  • Less space efficient – compare a new server with one of four years ago in terms of storage support, memory and cpu – a 1u box might easily be powerful enough to replace that 5u box from 6 years ago.
  • More time consuming to repair and fewer components can be ‘hot swapped’, newer servers are designed to be hot swap, the parts will be more readily available, try logging a call with your service provider for a system board for a Compaq server from 10 years ago, they will have one, but what’s the lead time? Days?
• How can I put this politely a focus on cost in the wrong way? Avoiding capex is one thing, but you need to ensure that you are not simply moving the expense, the budget around. I save £3000 by not upgrading my server, that means as IT, I have to:
  • Continue paying support for that operating system which might create an additional cost
  • Continue keeping additional components on site (which might be chargeable) to support that server
  • The server might easily be a support overhead -  small disks, not enough memory, be the cause of so many calls a year
  • The server might need cleaned up, brought in line with current standards, have the layered components, component firmware upgraded, it all takes time and organization
  • The server therefore might so easily generate 50% of it’s capital cost in support cost if you think about it, and what happens if we want additional local hard drives or memory – the cost of legacy components can be horrific.

Related posts:

1. Talking about everything in blade servers Earthweb Blade Servers are the hottest trend in datacenters today....
2. Outsourcing the IT amongst other things It will be interesting to see what comments or analysis...
3. Has the 1u undermined the other u servers? I was having a chat with a friend the other...
4. Asking the right questions is key I was talking with some friends, a project manager, a...
5. Windows 7 desktop deployments not a prioirty? Silicon have an article talking about when CIOs are deploying...

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