IT purchasers will soon find out exactly how serious server manufacturers are about developing a standard that allows customers to easily compare the energy efficiency of their products, after the US Environment Protection Agency (EPA) last week confirmed it plans to extend its Energy Star labeling scheme to cover enterprise servers.
In an open letter to server manufacturers Energy Star program manager Andrew Fanara revealed that the EPA "is initiating its process to develop an Energy Star specification for enterprise computer servers". He added that talks with manufacturers and datacentre managers had revealed "compelling evidence of the need for this specification", citing a recent survey that found 70 percent of datacentre operators regard power and cooling as the primary issue they now face.
The news will be welcomed by many server purchasers and environmental campaigners who have argued that an international standard for measuring server energy efficiency is the only way of resolving the current confusion arising from rival vendors using their own metrics to argue that their products are the most energy efficient available.
Many vendors have agreed that this is the case and have contributed to the development of the EPA's proposed measurement protocol, which is expected to form the basis of any Energy Star labeling system for servers.
However, while openly supporting the idea of a metric some industry figures have expressed reservations about whether a satisfactory metric can be developed given the wide number of different uses for servers and the multitude of different configurations that can be applied to them.
The proposed Server Energy Measurement Protocol, which was released late last year, has attempted to get round this problem by adopting existing performance benchmarks, such as load processing, and inviting manufacturers to measure energy consumption at a range of different workload levels.
This approach means that each server would generate an energy/performance graph or "power plot" that, according to the EPA, would allow server purchasers to "specify the performance levels at which they expect their servers to operate, based on their facility’s experience and then read off the power use (as a % of peak power use) at that server work load".
The proposal argues that these complex metrics would ensure customers are better equipped to pick the most efficient servers. "Once these power plots are developed by manufacturers as a matter of course, then server purchasers can use them to specify equipment," it explained. "For example, a server purchaser could rank by energy use of all servers that meet certain performance criteria at a given load level, and then consider more carefully the top 20% of servers by that metric."
Robin Murray, a consultant for DEFRA's market transformation programme who has worked on establishing the Energy Star system in the UK, argued that with vendors such as Intel, AMD, IBM, HP, Sun Microsystems and HP all involved in drawing up this protocol there is likely to be adopted as the basis for new Energy Star labels for the most energy efficient servers.
However, the EPA's letter does offer a get out clause to any vendors who remain unconvinced by the effectiveness of a standardised metric and the need for a labeling system.
Fanara not only reminds vendors that the Energy Star scheme is voluntary but also confirms that the "EPA will conduct an analysis to determine whether such a specification for servers is viable given current market dynamics, the availability and performance of energy-efficient designs, and the potential energy savings".
Critics have long argued that those vendors with the least energy efficient products may have a vested interest in stalling the development of a metric that would expose their poor performance. As such the EPA's declaration that it is to analyse whether the market could support Energy Star labels for servers represents the last chance for any opponents to stop its emergence.
Even if the EPA concludes the server market is ready for the Energy Star scheme opponents would still have plenty of opportunities to slow the formal launch of Energy Star labeled servers. The current measurement protocol will be refined throughout 2007, according to Murray, but it is far from the finished article.
For example, the EPA's proposal accepts that the use of different performance benchmarks could make comparisons between different "power plots" complicated. It also claims that with the current proposals only covering 1U and 2U servers a means of measuring power used by blade servers needs to be developed and that more data is needed on the typical datacentre workloads that should be measured.
The manner in which response times fall at lower workload levels further confuses matters with the EPA admitting that "in order to ensure a consistent comparison, explicit specifications may be necessary to establish response time constraints at various load levels". While the distinction between zero percent work load and idle/standby conditions could also undermine the accuracy of measurements.
Every one of these concerns will give opponents of a standardised metric ammunition with which to argue that the energy efficiency of rival servers is too complicated to be easily compared.
In short, it is crunch time for the server industry when it comes to energy efficiency. They can now either fall into line and support Energy Star as a valuable decision making tool for server purchasers and seek to iron out these issues as quickly as possible, or continue to pay lip service to standards while also confusing the market by promoting their own energy efficiency metrics.
Encouragingly, many of the big brand names look set to support the extension of Energy Star. Though it remains to be seen if they will follow up their involvement in developing the current protocol by side-lining their in-house metrics and helping to push through the EPA's standard.
The fact is servers are so complicated that no energy efficiency standard will ever be perfect – just as you could argue until the last glacier has disappeared over performance metrics, you can similarly argue over energy efficiency metrics.
But the EPA's graph-based proposal is a pretty effective compromise and hopefully vendors have at last realised it is more beneficial to both customers and the environment to have an agreed standard that may still need some work, rather than no standard at all.
New study predicts that despite pledges to tackle plastic waste, UK is still on track to throwaway a third more coffee cups by 2030
Parliament celebrates deep cuts in greenhouse gas emissions, as it marks Earth Hour 2018
From renewables 'revenue stacking' to new nuclear: Five key takeaways from Aurora Energy Research's annual get-together
Experts, CEOs, and policy wonks from across the UK energy sector descended on Oxford for Aurora Energy Research's Spring Forum this week - here's your need to know guide
Think tank welcomes Defra's post-Brexit farming subsidy plan, but says more focus needed on sustainable food production