On June 11, Intel sponsored an event in Santa Clara titled the Intel Technology Great Debate. I took notes then and again yesterday at the second event, held in San Francisco in conjunction with the Intel Developers' forum. This event yesterday brought in 30 or so new audience members whose interest in software development clearly dictated their perspective on the issues raised for data center energy efficiency. Both events were well attended and oversubscribed.
Here are some of my notes from June 11th:
Andrew Fanara spoke first. He cited a May 24 The Economist article. “Clearly there is frenzy around this. Good long term planning will get us out of it. Nothing less.” Lets avoid being bullied into bad decisions relevant to energy policy. World oil production is flat for 2005 – 2007. More costs will come down the pipeline to get it moving upwards again. Even with promising renewables, carbon based fuels (including 40% from oil) are still the largest share of our useable resources and energy efficiency is very important. To get the full range of technologies, CO2 might have to be $ 200 per ton, and it is currently less than 20% of that.
demand between 2008 and 2010 is equal to 10 new power plants, and data centers
are currently using 5% of production and expected to surpass airlines by 2020.
CO2 emissions are projected to quadruple from 170 MT (metric tons) to 670 MT. Carbon sequestration
is in limbo until our Congress makes some decisions – it has lots of promise
but won’t be realized anytime soon.
“The grid of today is not the grid of the
Customers, Boards, large investors, etc all want to know: how much energy do you use, so we can understand the risks? Risk and cost containment require measuring what you use. When will we have zero energy consuming buildings? It is worth thinking about data centers as a big part of the entire building consumption.
I then took a lot of notes about the debate topic (AC v DC) that in hindsight make very little sense because some of the team members argued for their competitor's position. So in retrospect the comments from Andrew are the most memorable part of the event.
At the August 18 debate there were two questions after an hour's summary of the various partners involved and their agendas. The two questions were:
Which is better: high or low density data centers? moderated by Matt
Eastman of IDC.
Verari, Dell, IDC Architects, Oracle, Microsoft were represented on the panels.
Which is better: data centers in concrete/brick buildings or storage containers?
First we discussed low density and high density attributes.
To begin with, how would the audience define low density and high density for the debate? We could use as examples: 6kw racks and under are low density, 7 – 12 kw are medium density and 13kw+ racks are high density. When density starts to drive design decisions it becomes high density.
Legacy cost allocations drive density: “he who goes to
higher density first pays less” is a quote from the low density team’s
slides. At some point in the future, 30k
racks will be low density, so any definition is a moving target. We need to be looking at how we charge business
units for power, said one panelist – that will drive the density decisions. Micosoft’s USA-based data center costs in the
Cost per megawatt or kw per rack are better metrics
than kw per square foot.
Shortest distance = less work to be done moving air, therefore
Shorter distance = more efficiency so density is GOOD!
But at the end of the voting, low density seemed more popular than high density because of the added costs to cool high density environments.The lifecycle for IT equipment is much shorter than that of the infrastructure so therefore, you have difficult meshing the two.
With respect to the 2nd debate, it was so congenial, with both sides admitting that the other had good points, and that new modular brick & mortar shell environments essentially offer many of the same benefits of the modularity in containers, that no clear consensus was reached.
Here is my recommendation to Laurie Wigle of Intel, who created both events, for Debate #3: Allow Henry Wong of Intel and Christian Beladi of Microsoft to moderate the panels so that the conversation between the contestants will be more evenly matched in personality. That means David and Matt from IDC (currently the moderators) will have to take opposite sides on the Tupperware v BAM teams and we'll see different dynamics emerge. Have a green day!