Thoughts on Climate Change
Interesting documentary on TV last night: ‘The Great Global Warming Swindle’. Superficially, it reminded me of Lomborg’s ‘the Skeptical Environmentalist’, but came over as much less coherent and consistent. I tend to the opinion that Climate Change (or the possibility of Climate Change) is really an economic problem than a scientific one.
Whatever. There’s a growing focus on the power and cooling needs of data centres. So, here’s a few data points that I find interesting:
- a typical corporate server has a comparable carbon footprint to a family car
- at above 1.5kW per sq m power input, you should put in a backup cooling system as if the primary system fails, even if the power in is cut instantly - not a nice scenario - the heat cannot be dumped fast enough to prevent equipment being destroyed. This can break the economics of some data centres
- a rack of blades takes in (and obviously pushes out) up to 20kW. Even though a rack takes up more than a square metre, that’s a lot of power coming and and heat to move out.
- most corporate servers are under-utilised: target utilisation is, of course dependant on what business applications are serving, but blades in particular typically run at 5% and often do nothing for the first few months of their lives
- virtualisation promises to improve utilisation (of large boxes), for the 1/10th to 2/3rds of the estate that’s used for development and testing, but most are not with it for production systems.
- Approaches such as Egenera’s architecture promise virtualisation like capabilities across blade based hardware platforms - this is powerful as it enables smaller incremental increases in processing power, but these have turned out to be expensive to manage.
- In large estates, 2-5% of servers that are still running have been removed according to the decommissioning process. A further 5-10% are idle. It’s just a question of finding them…...
Unsurprisingly, hardware and systems management tool vendors have spotted an opportunity here, and, as usual, if we replaced everything with their technology, there would be significant energy savings. But aside from the capital and project costs of re-platforming and testing the whole software estate, there are even more problems with the new technologies in tracking what’s going on in the estate than with current technology. I can just see a bonanza for the software vendors such as Oracle, Microsoft and IBM when it comes to true-up or audit time.
RSS