Power-aware scheduling problem has been a recent issue in cluster systems not only for operational cost due to electricity cost, but also for system reliability. In this paper, we provide SLA-based scheduling algorithms for bag-of-tasks applications with deadline constraints on power-aware cluster systems. The scheduling objective is to minimize power consumption as long as the system provides the service levels of users. A bag-of-tasks application should finish all the sub-tasks before the deadline as the service level. We provide the power-aware scheduling algorithms for both time-shared and space-shared resource sharing policies. The simulation results show that the proposed algorithms reduce much power consumption compared to static voltage schemes.

Cluster systems are prevalent infrastructures for offering e-services because of their cost-effectiveness. The objective of our research is to enhance their cost-effectiveness by reducing the minimum number of nodes to meet a given target performance. To achieve the objective, we propose a load balancing algorithm, the Nearest Underloaded algorithm (N algorithm). The N algorithm aims at quick solution of load imbalance caused by request departures while also preventing herd effect. The performance index in our evaluation is the xth percentile capacity which we define based on throughputs and the xth percentile response times. We measured the capacity of 8- to 16-node cluster systems under the N algorithm and existing Least-Loaded (LL) algorithms, which dispatch or transfer requests to the least-loaded node. We found that the N algorithm could achieve larger capacity or could achieve the target capacity with fewer nodes than LL algorithms could.

A parallel file system is normally used to support excessive file requests from parallel applications in a cluster system, whereas prefetching is useful for improving the file system performance. This paper proposes dynamic file prefetching scheme based on file access patterns, named table-comparison prefetching policy, that is particularly suitable for parallel scientific applications and multimedia web services in a VIA-based parallel file system. VIA relieves the communication overhead of traditional communication protocols, such as TCP/IP. The proposed policy introduces a table-comparison method to predict data for prefetching. In addition, it includes an algorithm to determine whether and when prefetching is performed using the current available I/O bandwidth. Experimental results confirmed that the use of the proposed prefetching policy in a VIA-based parallel file system produced a higher file system performance for various file access patterns.