A modelling approach for estimating energy consumption of NoSQL-based storage systems

A modelling approach for estimating energy consumption of NoSQL-based storage systems

연구 분야: Databases

학회: The Journal of Supercomputing

In large-scale data storage systems, NoSQL database management systems are commonly adopted for their capacity to provide better performance and availability. A key feature is the support for eventual consistency. Distinct consistency levels can be adopted, such that a designer may adjust the respective influence on performance and availability according to system requirements. Some research proposes techniques to estimate the impact of eventual consistency on performance and availability, but energy consumption is usually not taken into account. This paper presents a modeling approach based on stochastic Petri nets for estimating energy consumption of NoSQL-based systems utilizing quorum technique. The conceived model jointly considers performance and availability as well as the influence of consistency level, replication factor, and read and write operations. The proposed approach provides an additional design tool to evaluate system settings before the final implementation. Experimental results demonstrate replication factor and consistency level may impact energy usage more than 10%.