Issue 1, December 2008
FOCUS FEATURE MEASUREMENT
They have made an extraordinary claim, yet have not backed it with
any data, other than to say "trust us, we're Google"
PUE saying he is delighted that people are using it but acknowledging that it is one of several being developed. “It is great that people are using PUE although the use and uptake has been a little bit in front of its ability. The industry is trying to find a way to measure efficiency that is transportable between data centers, but it is not that easy. Because of the the metrics involved it is difficult to monitor and police what is going on. For example consultant PWC looked at companies that reported carbon emissions, of 30 companies, each was measuring and reporting it in a different way.
If they start measuring, that’s the first step and best practices exist to help with that. As an industry we haven’t done that well but data center productivity is now being looked at.”
OTHER METRICS
Smith recognises that the driving force is “let’s do something before it's mandated.” “People see energy as a big bad beast because the cost of energy is going up. We have this room which is sucking more power than the rest of building. Take a PUE graph – and it goes down. That comes off the bottom line. But it doesn’t mean you are using less power, all things being equal, over a 2 or 3 year period you are doing more work."
“We didn’t have the description of what to measure or where to measure so we developed PUE. We now have a task force working on making sure we get an even better understanding. It could be in the future that we can say a Tier IV data center should have a PUE of X, but we are not there yet. Even comparing two data centers within the same company is comparing different types of facility.”
It is well known that high utilisation, redundancy and resiliency all make for higher PUE numbers and while efficiency drives are welcome the standard by which to accurately measure efficiency has yet to be developed. Other metrics are being developed all the time by organisations such as The Uptime Institute and The BCS. And even PUE isn’t standing still. In a statement the Uptime Institute said: “We all understand that Google is unique in its ability to achieve such heroic feats as this, which may not be possible for the average corporate data center owner/operator. But at a minimum,
SCRUBBED NUMBERS
In his note on the subject Tier1’s Dan Golding wrote: “Google released some highly scrubbed datacenter efficiency numbers that T1R is scratching our heads over. Specifically, Google has announced that its well-placed, custom-designed, very large datacenters – quite new, too – are highly efficient. That part we believe. Specifically, Google's numbers track six of its mega-datacenters, which it has identified by letter: A, B, C, D, E and F. While our analyst team is busy tracking down which letters correspond to which tax-break-generous municipality (South Carolina, anyone?), we took a quick look at the numbers. Given the general parameters discussed above (these are massive, newer, custom facilities, operated very proficiently), some of the efficiency figures look reasonable. Most of the facilities, including two new facilities, clock in PUE measurements of about 1.25 on average; very good, but still in the realm of the believable, if they are situated in temperate climes with outside air cooling. One facility – and not a newer one – Datacenter B, on the other hand, reports PUEs as low as 1.13 on a quarterly basis. The accompanying diagrams show that Google is supposedly including all of the non-critical electrical loads it should – HVAC, PDUs, etc. So, has Google invented a magic datacenter? No, T1R doesn’t think so. We do think that Google's measurement methodology is likely very screwed up, at least in one of its facilities, which makes us skeptical of all of its figures. No particularly advanced or breakthrough technology is noted. In fact, cooling towers, not exactly a new technology, is cited as being the wonder invention that is resulting in these amazing numbers.
T1R has not seen substantiated operational PUE figures below 1.2 on an annualized basis and is highly skeptical that such facilities exist at current levels of technology, regardless of location. Google's release of these numbers, which do not seem to rely on any break-through technology, strains the limits of believability. While T1R does not think Google is being deliberately deceptive, Google's rather doctrinaire embrace of ‘green,’ combined with a now-familiar arrogance, may be causing it to embrace numbers that a more skeptical or humble firm would think required a second look.”
every data center operator ought to be able to measure his/her PUE and its improvement in relationship to other industry - and scale- comparable data center operations.” The Institute said it is working on an industry- wide accepted measurement/verification standard with the working title of SAVE (Site Analytics [to] Validate Efficiency). The methodology is intended to provide validated and independently certified energy performance figures and is expected to roll out in April 2009.
The conclusion is that PUE is a good starting point. That at the moment the PUE tells you were you are in isolation but not in comparison to other data centers and certainly not where you are going in power consumption terms. Who will be the first to declare a PUE number of 1? The real answer is who cares.
Bragging rights aside everyone now understands that efficiency is key to the industry's future. The game has changed and data centers must change rapidly.
CONSISTENT MESSAGES
Despite requests Google refused to be interviewed. But Urs Hölzle, the Vice President who designs the datacenters that power Google spoke with Focus contributor
Dave O’Hara (
www.greenm3.com).
We discussed Google's PUE of 1.15 which many have questioned, reports O’Hara. “Urs explained this is not possible in a general purpose data center, and this PUE was achieved by designing specifically for the Google HW in the facility. And, 0.05 of the PUE savings came from designing the data center specifically for Google's server design. As an example, Google designed the server fans to be efficient removing heat, and not waste energy moving excess air. Which Urs pointed out is a waste in many server designs, but important to Google.
It is ironic that if you chose to remove all fans from server HW and make it part of the cooling infrastructure, then your PUE would go up. Or can you count these fans as part of the IT load as it is if it is replacement for server HW? This is another detail I am curious to understand Google's approach.”
Even more ironically O’Hara also reported that “Urs expressed concern to be credible with his group’s PUE calculations and not be viewed as a marketing exercise.”
To be taken seriously in the rest of the data center industry Google and its PR machine need to remember that.
www.datacenterdynamics.com 25
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72