Posts tagged with Cloud Computing

If you follow AMD closely, you know that of late, we’ve been driving a pretty lively industry discussion around battery life metrics for laptops. The basic premise is this:  while the product description tags that describe laptops in your local retail outlet purport to tell you how long each computer’s battery life is, those figures are typically very overstated, as they are arrived at using a benchmark test that essentially measures battery life while the computer is idling. As our CMO puts it, that would be akin to measuring your car’s gas mileage based solely on highway miles. AMD is proposing that the industry adopt a “city/highway miles” rating system for consumer notebooks with two metrics, one that represents the battery life for the computer when at idle, and one that tells you how long the battery will last when you’re actually putting it through its paces. Others agree this approach makes sense.

As a server guy, it’s my contention that the exact same principles apply in the datacenter, and better metrics are needed to help users, especially cloud users.

Thermal Design Point (TDP) is a metric which measures the maximum amount of power the cooling system in a computer is required to dissipate and is used by system designers. But we at AMD believe that TDP is almost irrelevant in helping IT staff plan a realistic power budget for their datacenter because it tells you absolutely nothing about how much power a chip will typically draw when running under normal loads.  There’s also an issue with using the TDP metric correctly.  AMD discloses the maximum TDP; after all, the maximum number is what a system designer needs to build their server.

In 2007, we introduced a new power consumption measurement, called “Average CPU Power.” While no measure is perfect, ACP is designed to give customers a more accurate idea of the power consumed by the processor based on measuring the power consumption of a number of different workloads at the processor while operating in a more typical thermal environment. As we said when we launched ACP, “a processor with a 115W TDP may not break the 70W mark under extremely high workloads – just like a car with a 150MPH speedometer rarely hits speeds above 90MPH. Some customers were unnecessarily limiting the amount of growth within their server racks based on an over-estimated power budget and potentially sacrificing data center efficiency.

The most important reason why I think ACP is more relevant than ever before is interest in cloud computing is greater than ever. More and more users are looking for answers to questions about how cloud computing impacts them and their business. ACP numbers can be very indicative of what customers would see in a cloud computing environment. In our experience, the cloud environment encompasses balanced workloads with more emphasis on I/O and virtualization.  And in contrast to the “worst case possible” scenario measured by TDP, cloud servers are rarely completely idle and can typically operate between 10 – 50% of their maximum, thus providing headroom to scale if needed.

What’s my point? ACP is a fantastic metric and really the most useful processor power draw metric out there for IT and facility managers to leverage for their power budgeting calculations and models when planning a real world datacenter (again, we might be biased since we developed it.) Self-congratulating for ACP aside, the reality is nothing beats measuring the power drawn by a server at the wall. However, it appears that ACP, introduced well before cloud computing became all the rage, was actually rather prescient.  Suddenly having our head in the clouds doesn’t sound so bad, does it?

Brent Kerby is a Senior Product Marketing Manager at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.