You probably heard Sun's announcement this month of the UltraSparc T1 processor (See story). The new chip has many features to spark (pardon the pun) interest on its own merits, but at its launch Sun chose to focus on just one aspect: its power consumption The UltraSparc T1 pulls a frugal 70 watts, with Sun claiming that to be about half of what other CPUs eat up.
To drive that point home, Sun gathered a handful of distinguished guests, experts in environmental issues, at the Presidio in San Francisco. The panel was invited to discuss a rather controversial topic: the electricity consumption of computer equipment and its impact on the environment. See a link with the UltraSparc T1 yet?
If you missed the event, you can catch up with the event Webcast which includes a short stage-setting session from Sun Chairman and CEO Scott McNealy followed by a rather captivating panel discussion.
On that Webcast, you'll occasionally hear McNealy and Greg Papadopoulos, Sun's executive vice president and CTO, take shots at power-hog servers from competitors and bang the Sun thin-client drums (yes, still). The discussion is well worth listening to, however, because it provides some good facts about why reducing the electricity consumption in the datacenter is important and why you should care.
What does this have to do with storage? Absolutely nothing. I believe the word "storage" wasn't even mentioned once at that event, which is exactly the point I want to make.
Reducing power consumption of CPUs and servers is good, but what about storage arrays and other storage devices? Although it's true that many servers -- to use McNealy's verbiage -- behave like "space heaters," the amount of electricity they consume and the heat they generate pale when compared to some storage devices.
For example, according to Sun specs, a StorEdge 9990 with just one control frame plus one array consumes about 16kVA (kilovolt amperes) or 13kW (kilowatts). Ouch!
That's not to say that Sun's storage devices are the only electricity hogs; if you check the specs of high-end storage solutions from other vendors, you'll find similar numbers, although comparing apples to apples is not always easy.
For example, how do you measure the power efficiency of a Hitachi Data Systems (Profile, Products, Articles) TagmaStor versus solutions with a completely different architecture, such as an EMC (Profile, Products, Articles) DMX 3000 or an IBM (Profile, Products, Articles) Shark? Good luck on that one. If you think that comparing usual criteria such as performance, capacity, and scalability between heterogeneous storage solutions is a tough call, adding power consumption can really drive your analysis up the wall.
Perhaps we are spoiled because when shopping for household appliances we can easily check the EnergyGuide label for information such as the device's electricity consumption, the estimated yearly cost of use, and how it compares with other models.
Unfortunately, there isn't an EnergyGuide label for servers -- nor for storage, I might add. We absolutely need similar information for servers, storage, and other IT gear, especially considering that we're facing higher electricity costs in the foreseeable future. Moreover, a recent tendency to keep more data online rather than on tapes will cause the number of spinning disks, and therefore the electricity used in your datacenter, to increase.
Furthermore, although there is disagreement among scientists, some estimate that using more energy can have catastrophic environmental fallbacks. Katrina docet.
Should we trust the wisdom of our congressmen to create a regulatory noose to measure power consumption in the datacenter?