Although there is wide agreement that backfilling produces significant benefits in scheduling of parallel jobs, there is no clear consensus on which backfilling strategy is preferable - should conservative backfilling be used or the more aggressive EASY backfilling scheme. Using tracebased simulation, we show that if performance is viewed within various job categories based on their width (processor request size) and length (job duration), some consistent trends may be observed. Using insights gleaned by the characterization, we develop a selective reservation strategy for backfill scheduling. We demonstrate that the new scheme is better than both conservative and aggressive backfilling. We also consider the issue of fairness in job scheduling and develop a new quantitative approach to its characterization. We show that the newly proposed schemes are also comparable or better than aggressive backfilling with respect to the fairness criterion.

srinivas,kettimut,subraman,saday¢ Although there is wide agreement that backfilling produces significant benefits in scheduling of parallel jobs, there is no clear consensus on which backfilling strategy is preferable e.g. should conservative backfilling be used or the more aggressive EASY backfilling scheme; should a First-Come First-Served(FCFS) queue-priority policy be used, or some other such as Shortest job First(SF) or eXpansion Factor(XF); In this paper, we use trace-based simulation to address these questions and glean new insights into the characteristics of backfilling strategies for job scheduling. We show that by viewing performance in terms of slowdowns and turnaround times of jobs within various categories based on their width (processor request size), length (job duration) and accuracy of the user’s estimate of run time, some consistent trends may be observed. 1

The utilization of parallel computers depends on how jobs are packed together: if the jobs are not packed tightly, resources are lost due to fragmentation.

In this paper, we evaluate the performance implications of using a buddy scheme for contiguous node allocation, in conjunction with a backfilling job scheduler for clusters. When a contiguous node allocation strategy is used, there is a trade-off between improved run-time of jobs (due to reduced link contention and lower communication overhead) and increased wait-time of jobs (due to external fragmentation of the processor system). Using trace-based simulation, a buddy strategy for contiguous node allocation is shown to be unattractive compared to the standard noncontiguous allocation strategy used in all production job schedulers. A simple but effective scheme for selective buddy allocation is then proposed, that is shown to perform better than non-contiguous allocation. 1

The utilization of parallel computers depends on how jobs are packed together: if the jobs are not packed tightly, resources are lost due to fragmentation. The problem is that the goal of high utilization may conflict with goals of fairness or even progress for all jobs. The common solution is to use backfilling, which combines a reservation for the first job in the interest of progress with packing of later jobs to fill in holes and increase utilization. However, backfilling considers the queued jobs one at a time, and thus might miss better packing opportunities. We propose the use of dynamic programming to find the best packing possible given the current composition of the queue, thus maximizing the utilization on every scheduling step. Simulations of this algorithm, called LOS (Lookahead Optimizing Scheduler), using trace files from several IBM SP parallel systems, show that LOS indeed improves utilization, and thereby reduces the mean response time and mean slowdown of all jobs. Moreover, it is actually possible to limit the lookahead depth to about 50 jobs and still achieve essentially the same results. Finally, we experimented with selecting among alternative sets of jobs that achieve the same utilization. Surprising results indicate that choosing the set at the head of the queue does not necessarily guarantee best performance. Instead, repeatedly selecting the set with the maximal overall expected slowdown boosts performance when compared to all other alternatives checked.

Abstract. Application scheduling plays an important role in high-performance cluster computing. Application scheduling can be classified as job scheduling and task scheduling. This paper presents a survey on the software tools for the graphbased scheduling on cluster systems with the focus on task scheduling. The tasks of a parallel or distributed application can be properly scheduled onto multi-processors in order to optimize the performance of the program (e.g., execution time or resource utilization). In general, scheduling algorithms are designed based on the notion of task graph that represents the relationship of parallel tasks. The scheduling algorithms map the nodes of a graph to the processors in order to minimize overall execution time. Although many scheduling algorithms have been proposed in the literature, surprisingly not many practical tools can be found in practical use. After discussing the fundamental scheduling techniques, we propose a framework and taxonomy for the scheduling tools on clusters. Using this framework, the features of existing scheduling tools are analyzed and compared. We also discuss the important issues in improving the usability of the scheduling tools.

Resource management systems (RMS) for networks of workstations exploit unused computing capacities by automatically assigning jobs to idle workstations. Until now, the WINNER RMS developed by our group only provides support for interactive jobs. In this paper, an approach to seamlessly integrate a queueing system for batch jobs is presented. The basic design decisions, the queueing system's functionality, and the benefits of its use within our workstation environment are described.

A taxonomy of application scheduling tools for high performance

A number of applications comprise of several small independent and homogeneous tasks that need to be executed. However, most supercomputer centers enforce a restriction on the number of jobs a single user can submit to the cluster at any time in order to ensure fairness to the other submitted jobs. This forces the users to combine these independent homogeneous tasks into a single parallel Parameter Sweep Application (PSA). In this paper we propose a new and efficient scheme, termed as Opportune Job Shredding, which allows Supercomputer Centers to take advantage of the independence of the component tasks of Parameter Sweep Applications, without affecting the other submitted jobs significantly. We also propose an extension of the previously proposed ”Multiple Simultaneous Requests ” scheme combining it with the Opportune Job Shredding scheme allowing Parameter Sweep Applications to be executed in parts on remote clusters.

Resource management systems (RMS) for networks of workstations exploit unused computing capacities by automatically assigning jobs to idle workstations. Until now, the WINNER RMS developed by our group only provides support for interactive jobs. In this paper, an approach to seamlessly integrate a queueing system for batch jobs is presented. The basic design decisions, the queueing system's functionality, and the benefits of its use within our workstation environment are described. Keywords: networks of workstations, resource management, batch queueing system 1 Introduction Networks of workstations (NOWs) are by now ubiquitous general purpose computing platforms. Since console users of workstations hardly utilize the processing capabilities of their machines (e.g. while editing text, reading mail, browsing the web, or being physically absent), the idle time of workstations is often as high as 95% [10]. Simultaneously, there is high demand for computing power driven by applications f...