Towards green information retrieval: studying the suitability of queueing theory in reducing power consumption

Many efforts are being dedicated to reduce the power consumed by large data centres. However, less research has focused on developing environmentally-friendly search engines. This paper contributes to Green Information Retrieval by defining a mathematical model to automatically vary the number of powered-on servers of a replicated search engine, based on the load of the system. Following general-purpose data centres, we base our approach in Queueing Theory, in order to study its behaviour when being applied into a search engine. Results show how our model can achieve energy savings by maintaining excellent values for the latency, and it also allows us to study the shortcomings of Queueing Theory into Information Retrieval.