No abstract

No abstract

Abstract. Backfill is a technique in which lower priority jobs requiring fewer resources are initiated before one or more currently waiting higher priority jobs requiring as yet unavailable resources. Processors are frequently the resource involved and the purpose of backfilling is to increase system utilization and reduce average wait time. Generally, a scheduler backfills when the user-specified run times indicate that executing the lower priority jobs will not delay the anticipated initiation of the higher priority jobs. This paper explores the possibility of using a relaxed backfill strategy in which the lower priority jobs are initiated as long as they do not delay the highest priority job too much. A simulator was developed to model this approach; it uses a parameter ω to control the length of the acceptable delay as a factor times the wait time of the highest priority job. Experiments were performed for a range of ω values with both user-estimated run times and actual run times using workload data from two parallel systems, a Cray T3E and an SGI Origin 3800. For these workloads, overall average job wait time typically decreases as ω increases and use of user-estimated run times is superior to use of actual run times. More experiments must be performed to determine the generality of these results. Scheduling PoliciesMany practical job scheduling policies, whether for uniprocessor or multiprocessor systems, incorporate the notion of job priority. Perhaps the simplest example of a priority scheme is setting a job's priority to elapsed time in the queue; if this priority scheme is used to dictate the order of job initiation, then a "first-come, first-served" (FCFS) policy results. Other, more elaborate schemes based on the number of processors requested and user estimates of run time are, of course, possible. A second important concept involves how to use the resulting prioritized list of jobs. If, when a job completes, the prioritized list is searched for the first job that will run using the available number of processors, then a "first-fit, firstserved" (FFFS) policy results; there are also "best-fit, first-served" (BFFS) (to fit the available number of processors the tightest) and "worst-fit, first-served"

Often, images or datasets have to be compared, to facilitate choices of visualization and simulation parameters respectively. Common comparison techniques include side-by-side viewing and juxtaposition, in order to facilitate visual verification of verisimilitude. In this paper, we propose quantitative techniques which accentuate differences in images and datasets. The comparison is enabled through a collection of partial metrics which, essentially, measure the lack of correlation between the datasets or images being compared. That is, they attempt to expose and measure the extent of the inherent structures in the difference between images or datasets. Besides yielding numerical attributes, the metrics also produce images, which can visually highlight differences. Our metrics are simple to compute and operate in the spatial domain. We demonstrate the effectiveness of our metrics through examples for comparing images and datasets.