Energy Efficiency Research On An ARMv7 Cluster

Traditionally, datacenters employ x86-based server nodes with power usage of tens of hundreds of Watts. Power cost is becoming the main part of the total own cost of the cluster facilities. Future exascale supercomputers will be strongly constrained by their power requirements. With the popularity of smart mobile phone, the performance of the cluster of ARM-based multi nodes is better than that of an x86-based architecture high performance server. The use of high performance ARM-based processor to build low-power cluster becomes very attractive research problem.Taking the energy consumption of ARM cluster as the object of study, the effect of ARM-based CPU?I/O and workloads on the energy efficiency of ARM cluster is analyzed in detail, the running time of the tasks in the cluster is decomposed, and the CPU time of the cluster can be assigned to the task with the scale expansion is the key metric. An energy efficiency model of ARM cluster is proposed in the Map Reduce distributed computing framework depend on this. According to the characteristics of ARM hardware, the energy consumption performance of ARM cluster is evaluated under different computing and I/O performance requirements based on the energy efficiency model of ARM cluster.In terms of this theory model, design and build a lightweight ARMv7-based cluster, using different types of Map Reduce workloads to test the ARM-based cluster can analyses the influences of the CPU/?I/O and the types of the workloads on the ARM-based cluster energy efficiency. The correctness of the ARM cluster performance and power consumption model is verified, which provides a reference for the design and deployment of the cluster based on ARM. Experiments show that, with the application of the computation intensive the ARMv7 cluster will provide better calculation and energy efficiency. The number of ARM nodes must be controlled within a certain range to maintain its energy efficiency advantages while building a low energy consumption cluster using ARMv7 architecture.