Supersizing the data center. Not good
Two papers about power and its excesses
Off we shuffle to the Reg Library for our occasional trawl through the whitepaper section. This week we present a couple of reports that shed light on heat and power in the data center. It may be too late for the data centers you are operating today, but we think both paper will stimulate thinking for those planning new data centers. Here we go:
This IDC whitepaper, sponsored by Dell and Intel, is based on a survey of 2000 datacenter owners in 14 Asia-Pac countries. From the findings, authors Avneesh Saxena and Daphne Chung, tease out the 'datacenter pain points' that operators suffer.
Most respondents think their datacenters are reaching end-of-life: they were built a mere six to 10 years ago. Many find it harder to cope with the higher power densities of today's servers; with cooling those servers; and with containing their power bills. Almost all respondents reported difficulty in establishing how long the current datacenter would be running. They also indicated a "critical need to revamp or retrofit existing datacenters in order to accommodate additional servers or even keep the existing one running smoothly".
Today's datacenters are very inefficient, typically with a lot of redundancy built in, to allow for unexpected needs. This is expensive. To cope, datacenter operators are looking to virtualization, consolidation and better systems management tools, IDC reports. The white paper ends with an overview of Dell's Poweredge 11g servers, powered by Intel's Nehalem chips, and the vendor's system management software.
This is a good paper, long on facts, short on vendor guff.
Taking up the redundancy baton, APC wades in with a typically thorough paper on the costs associated with oversizing data centers. According to APC, oversizing is the "single largest avoidable cost" associated with typical data center and network room infrastructure.
Retro-fitting costs are so substantial, that oversizing is an entirely logical approach to take, APC says. But the company pinpoints defects in the planning process that makes datacenter operators over estimate growth needs by as much as three times. [It bases this claim on data from actual installations.]
Under-utilization means unnecessary maintenance cost and electricity. APC calls for a modular approach to building infrastructure that can help datacenter owners accommodate unforeseen growth without supersizing. ®
You can buy gas powered fridges, they burn gas and somehow this drives a heat exchanger
Could you use the servers heat to power its aircon?
The hotter it gets the more heat it pumps away
Another idea is SOB, or System On a Brick, all your servers would be on the outside wall, if you needed more you could pop down B&Q
Web 3.0 people could choose wireless mortar
APC can provide this kind of modularization..
surprising, isn't it? *G*