The Myth of Performance in Green IT

Performance per watt is often taken as a measure of the energy efficiency of a given system. It is just a view of the computation that can be performed per energy unit.

The question, then, is how do we measure  performance?

The performance metrics used depend on the definition; reasonable measures of performance are FLOPS, MIPS or the results of benchmarks. Computing energy consumption is sometimes also measured by reporting the energy required to run a particular benchmark, for instance . Energy consumption figures for a standard workload may make it easier to judge the effect of an improvement in energy efficiency, just as the use of L/100km is easier than reciprocal measures when judging impact of a car gas consumption. Performance (in operations/second) per watt can also be written as operations/watt-second, or operations/joule, since 1 watt = 1 joule/second.

However, the Megaherth Mith arises here obscuring the easiness of performance (and energy efficiency) measuring.  This myth refers to the error of using clock rate to compare the performance of different CPUs. While clock rates are a valid way of comparing the performance of different speeds of the same model and type of processor, other factors such as pipelines and instruction sets can greatly affect the performance when considering different processors.

Comparisons between different types of processors are difficult because performance varies depending on the type of task. Then, is quite complicated to have a general view of the energy efficiency of a system. This is a long unsolved question often dealt with in undergrad  computer architecture courses.

Although the question still remains, it is important to keep these considerations in mind when analyzing the application of Green It solutions and the comparison of different products/strategies.

New IEEE Computer Magazine Column: “Green IT”

Bamboo Forest by Roger Walch

The IEEE Computer magazine, one of the most important in the ICT field, has created a new column devoted to the emerging area of “Green IT” (aka “Green Computing”). The first installment calls for contributions that provide insights about research works, trends and best practices in Green IT.

Read the rest of this entry »

Tags: ,