The advent of cloud computing has brought many exciting changes to companies’ IT strategies. One aspect of the cloud that is frequently overlooked, however, is energy efficiency. On the face of it, one might expect cloud computing to be more energy efficient than the alternative. But is it really?
Let’s take a quick look at the three drivers behind increased energy efficiency in cloud environments.
First and most obvious is economies of scale. It’s not rocket science to understand that fixed costs are best allocated among a greater quantity to bring down per-unit cost. Similarly, conducting a benchmarking exercise to measure Power Usage Effectiveness entails significant fixed costs in devoting resources to counting equipment and measuring individual devices’ power consumption. There are certainly economies of scale to be gained in doing this for a larger datacenter than for a smaller one.
The second driver of energy efficiency in cloud environments results from the abstraction of the physical and virtual layers in the cloud. A single physical server running multiple server images will obviate the additional power load from purchasing additional physical servers. Also, if a virtualized environment incorporates redundant server images on different physical boxes, then individual boxes do not need multiple power supplies. The failure of one machine becomes a non-issue when redundancy is built in.
Finally, a datacenter serving cloud clients will have more users from more disparate places, each with different needs. This means that system loads will be more evenly spread throughout each day (and night), which enables the datacenter to average higher system loads and thus more efficient utilization of equipment. Everest Group research shows that individual servers in a cloud datacenter experience three to four times the average load of those in an in-house datacenter.
By now it should be clear that a large cloud datacenter has distinct energy efficiency advantages over a smaller, in-house datacenter. But there are corresponding energy drawbacks to cloud migration that may not be immediately apparent. First, as processing and storage shift to the cloud, energy usage increases. This is primarily from the routers transporting the data over the public Internet; their power use increases with throughput and frequency of accessing remotely stored data.
Also, in a SaaS, PaaS, or simple cloud storage scenario, frequent data access can cause data transport alone to account for around 60 percent of the overall power used in storing, retrieving, processing, and displaying information. At this point, the efficiency advantages gained by the three drivers cited above may be lost due to the extra power required to move the data between the user and the cloud datacenter in which it is stored or processed.
It is true that migration to the cloud can yield significant gains in energy efficiency for certain applications. However, for applications involving high transaction volumes, an in-house data center can provide better energy efficiency.
As power prices become increasingly important in determining data center operating costs, energy efficiency will play a greater role in companies’ cloud strategies.