Evaluating load balancing policies for performance and energy-efficiency

Abstract

Nowadays, more and more increasingly hard computations are performed in challenging fields like weather forecasting, oil and gas exploration, and cryptanalysis. Many of such computations can be implemented using a computer cluster with a large number of servers. Incoming computation requests are then, via a so-called load balancing policy, distributed over the servers to ensure optimal performance. Additionally, being able to switch-off some servers during low period of workload, gives potential to reduced energy consumption. Therefore, load balancing forms, albeit indirectly, a trade-off between performance and energy consumption. In this paper, we introduce a syntax for load-balancing policies to dynamically select a server for each request based on relevant criteria, including the number of jobs queued in servers, power states of servers, and transition delays between power states of servers. To evaluate many policies, we implement two load balancers in: (i) iDSL, a language and tool-chain for evaluating service-oriented systems, and (ii) a simulation framework in AnyLogic. Both implementations are successfully validated by comparison of the results.Comment: In Proceedings QAPL'16, arXiv:1610.0769

Specification of Data Centre Power Management Strategies

“…However, jobs can only be scheduled to servers in the idle and processing power states. In order to achieve this, we used the module extension proposed in [16].…”

“…e module extends PM functionality and introduces an interface to describe multiple strategies. Our recent work [16] introduced a module to this simulation framework, that performs PM depending on how jobs are scheduled among the servers. e approach in this paper is much more general, in the sense that it allows to use many more state variables than only queue size and time spent idling.…”