August and the world is becoming Greener!
I write this blog entry from my workplace in Edinburgh at the start of the 2007 International Festival and in the midst of the Festival Fringe which accompanies this event. Edinburgh becomes a city with twice the population at this time of year, the Scottish schools go back in a week or so, and so everyone is in town. The International Festival is holding its own on an increasingly crowded international stage, where there seem to be non-stop Festivals held during the summer months. But Edinburgh has one chief advantage over many, it's a beautiful city with much of historic relevance to see and sufficiently different to be noticeable in this increasingly homogeneous world. A great backdrop for the many ingenious ways people find of having fun.But why, I hear you ask, is this relevant? No particularly good reason, other than I fought my way through the crowds at lunchtime to run an errand and just enjoyed the fact that there were so many here already, Edinburgh's Festivals (for they include the Book; Film and just finished Jazz instances) run for the whole month of August, and the fact that it was warm outside!
And that brought me back to Green IT. In the past two months, I have attended the EC's 2nd meeting aimed at developing a Code of Conduct for Data Centres; listened to a particularly well articulated argument for change in the Data Centre from BCS colleague, Liam Newcombe, and enjoyed participating in the closing panel session at Information Age's 2007 Future of the Data Centre Conference, a sort of unofficial end to the London conference season before the hols. The strong thread running through these events was the increasing concern at power usage (abusage) in the Data Centre, which when coupled with the historic poor utilisation of compute resources makes for depressing reading. It may well be that our industry is responsible for generating the same level of carbon emissions as aviation, but at least the planes carry more than an average 15% of customers!
This is a story which is going to run for a while, if you haven't already read it elsewhere, you heard it here first! The Green Grid is a supplier led consortium which is looking to bring some order and understanding to the picture, they've recently published some measures for data centre productivity, rather more related to computing function and power consumption, than delivered value. But, hey, I am not sure that we know how to measure the latter anyway.
We've also seen a rash of collapsing information infrastructures being announced, collapsing in the way that matter falls into itself towards a black hole. HP and IBM have announced eye watering initiatives to massively reduce their business support infrastructures, the former using Virtualisation and standardisation; the latter a combination of this and the resurrection of the mainframe. You can guarantee that they'll be round to help you with your plans soon.
This approach certainly seems worthwhile and if that means we can switch off older, less efficient systems, then all to the good. If it also means that we can get more productivity from our servers then even better. But there is one problem that is not so easy to deal with, and I am not talking about finding the owner of a system in the corner which has been running for years but nobody remembers what it does. The problem that does concern me is how to migrate legacy applications into virtualised/distributed environments.
If you are one of the many users of mainstream commercial application bundles from MS, Oracle, SAP, SAS, and so on, you can relax, in that these organisations are running fast to provide "service enabled" application stacks which will help in the process of migration, isolating themselves from specific hardware implementation and revealing themselves as fashionable web services. But what about the myriads of other legacy applications which are bound to a physical hardware infrastructure and may not be wedded to an abstract, virtualised infrastructure? Where do you get help for that?
There's no ready answer, other than to say your best bet is simple virtualisation at the server level and hope that the VM is compatible with your application software. There is some room for optimism with VMWare and Xen, but it's almost certainly a suck it and see equation! Still, the good news is that VM overhead is typically in the single percentage realm and process performance advances are typically in the 10-20% for each generation, so you should win, you should be able to turn off some old equipment and run those apps alongside other, newer stuff - yielding an impressive advance in productivity all at the same time!
Of course, the proper thing to do is invest in developing an infrastructure, a service-oriented architecture and assemble your applications from web services. But that's another story and one which you almost certainly don't have the time, or the budget for today. Better get those operational costs down first and fight off the Financial Director's eagle eyes when prospective savings emerge. You'll need that money to do the job properly! Good luck!!
Labels: ; multi-core; power management, ; SOA, Edinburgh Festival, Green IT, Green IT; Semantic Web; Roadmapping, multi-core; power management, virtualisation, Web Services
posted by Ian Osborne, Grid Man Now! @ 5:35 AM
0 comments
