As it pushes more of its software into the cloud for customer delivery, Microsoft has a vested interest in ensuring that the data center infrastruture behind those servcies are as energy-efficient and sustainable as possible from an operational standpoint. Naturally, it is trying creative strategies to achieve this, and the company just decided to share some of them in a new white paper, "Microsoft's Top 10 Business Practices for Environmentally Sustainable Data Centers."
Here's a snapshot of those ideas, which come from the company's Global Foundation Services team:
1. Offer internal incentives. Specifically, Microsoft has created a chargeback structure that makes it in business managers' interest to buy lower powered servers, increase utilization and differ purchases. What's more, data center managers aren't just paid bonuses on uptime, they are recognized for improving energy efficiency as measured by power usage effectiveness ratio (PUE). That's the metric that expresses the amount of cooling and power infrastructure it takes to run your IT equipment. Microsoft has an average PUE of 1.4 for its cloud infrastructure in 2011; its goal by the end of 2012 is to hit an average of 1.125 PUE and use 30 percent to 50 percent less energy that traditional data centers.
2. Strive for optimal utilization. This means, frankly, that IT teams need to become much more acquainted with power and cooling costs, so that IT investments make the best use of what is already available. Microsoft notes: "If only 50 percent of a data center's power capacity is used, then highly expensive capacity is stranded in the uninterruptible power supplies (UPSs), generators, chillers and so on. ... Stranding capacity will also force organizations to build additional data centers sooner than necessary."
3. Invest in virtualization. I'm sure this revelation doesn't really surprise anyone reading this. The point that Microsoft makes here, though, is that underutilized servers continue to be a challenge for data center managers. It is consolidating onto virtual machines as quickly as possible. "A server running virtualization will often need more memory to support multiple virtual machines and there is small software overhead for virtualization. However, the overall value proposition measured in terms of work done per cost and per watt is much better than the dedicated underutilized physical serer case," Microsoft writes.
4. Seek compliance. The theory here is that quality is closely linked to earning these credentials, plus you get third-party validation of what your team's operational approach. This is mostly about security, but sustainability is implied. Among the certifications that Microsoft holds are ISO/IEC 27001:2005 and SSAE 16/ISAE 3402 SOC 1, 2 and 3.
5. Enforce proper change management policies. That means having a systematic approach to understanding how a planned infrastructure or process change might impact energy utilization, as just another risk to consider.
6. Document application workload and behavior. This will ensure that hardware choices (such as memory configurations) are more precisely configurated for actual (as opposed to guess-timated) needs.
7. Right-size server and network platforms. How many more slots and I/O ports do you REALLY need? Think before you are tempted to buy bigger when leaner might be the better strategy for a hardware procurement. This is a theme that I have heard eBay's Dean Nelson advocate, as well. His story of discovering that eBay used to use more than 200 different server configurations really sticks with me. As part of the company's Project Mercury green data center plan, it has reduced that number to two base configurations.
8. Evaluate and test servers for performance, power and total cost of ownership. The list of requirements for anything that Microsoft short-lists includes energy costs, and it favors using more efficient power supplies (even if they cost more) to balance out electricity expenses over time.
9. Standardize infrastructure. This is an echo of No. 7, and the eBay strategy I've already mentioned. Microsoft has also reduced the number of custom hardware configurations that its business units can buy. It has also modeled this around a 12 to 18 month cycle, so that it doesn't keep changing.
10. Take advantage of competitive bids. It bears repeating: Microsoft's total cost of ownership metrics for buying hardware consider the lifetime energy costs and data center allocations costs that are associated with the equipment -- in context. There are plenty of vendors trying to be as "green" as possible. Use that to your advantage.