The third annual event took place on Thursday Oct 14th, 2010 and it was really successful. It was a full audience and the feedback on the presentations throughout the day was very positive. This resulted from the maturity of the data presented. Instead of predicting results, presenters offered up examples of their findings - this provided a lot of content granularity. The topics differed from 2009 as they went from what we should do, to what we have done, from the prescriptive to the descriptive. Moreover, there was a sustained effort to discuss the costs: what was spent and how much was saved. I can't cover all the presentations because I didn't see them all, and I am particularly sorry that I missed most of Chill-Off @ cooling presentations. But the presentations will be online. Stay tuned!
Dave Stevens, CTO of Brocade, the hosting company for this year's event said that there will be 15B devices connected to the Internet by 2015, and there will be 35 Zetabytes of data stored by 2020. Those numbers are too big to really digest, aren't they? Significantly, there will be 4x more data stored outside of data centers than inside.
One of the presentations that I enjoyed hearing was an update about outside air economizers. In a panel led by Bill Tschudi, the ASHRAE whitepaper on the topic was politely shredded. In every instance of particulate matter, carbon, sulfate, nitrate, chloride, and black carbon, the air economizers were sufficiently robust and servers were unharmed when a MERV 7, 11 or 14 filter was installed. In practice, with a MERV 11 installed, the air quality was equal to that of HVACs, and where MERV 14s were installed, the air quality was better than more traditional air handling units. (MERV is an industry standard rating, so it can be used to compare filters made by different companies.) That should put the issue to rest.
In 2009, it was estimated that the USA had 1.35 million server closets (IDC). Joyce Dickerson provided this information in her panel presentation that I moderated, in the context of the 200 or so that have been identified at Stanford University - and she conjectures the actual number might be 2x that. Joyce said that her own analysis showed that they are not all inefficicent - and may run more efficiently at cost per watt than the mission critical data center. And, she continued, why consolidate if the efficiencies aren't improved and the business units don't want to be part of the larger data center in the first place? Perhaps consolidation is not the better alternative. Also on the same panel was Victor Garcia, Senior Manager of Facilities at Brocade. Victor had an entirely different scenario because Brocade, in their data center consolidation, had their own Brocade networking gear that made significant contributions to energy efficiency. They were able to save money with flat floors using no ramps, cutouts, etc., and eight foot custom built racks since they didn't use raised floors.
It became apparent to me that the case studies could be divided as follows: those that saved significant energy with better PUEs, and those where the PUE improvement was not particularly noteworthy but the waste heat was re-used. Both are viable alternatives, although 'using less' tends to rank higher than re-use on most sustainability hierarchies. Here is where you will want to look for the uploaded presentations next week. Rich Miller published photos of the event. Look for more to come from Kevin Heslin. I'll post when they are up. Have a green day!