V. K. Naik, S. K. Setia, and M. S. Squillante. Scheduling of large scientific applications on distributed memory multiprocessor systems. In Proc. SIAM Conf. Par. Proc. Sci. Comp., 913--922, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... the available processors into equal sized partitions has been studied with analytic models [Sev89, MEB91, ST93, SST93] compared with adaptive policies [Sev89, MEB91, GST91, RSD 94, ST93] compared with timeslicing policies [SST93] and compared with both adaptive and dynamic policies [NSS93b, NSS93a, NSS97] The general conclusion is that fixed partitioning can perform well if the workload is wellunderstood and jobs tend to have low variation in parallelism and service times. In many cases, a particular size of fixed partitioning outperforms all other policies for a particular system ....

....the newly freed processors are divided evenly between the jobs in the queue in FCFS order, subject to no job exceeding its maximum parallelism. ASP was found to outperform two variations of the pws policy [GST91] and a fixed partitioning policy. A variation on ASP is studied by Naik et al. NSS93b] where the allocation size is also limited by a system wide minimum. This policy performed nearly as well as fixed policies at particular loads, but was more consistent across different configurations. A dynamic policy allowing limited pre emption was also studied, but it tended to perform ....

[Article contains additional citation context not shown here]

Vjay K. Naik, Sanjeev K. Setia, and Mark S. Squillante. Scheduling of large scientific applications on distributed memory multiprocessor systems. In Linda R. Petzold Richard F. Sincovec, David E. Keyes, Michael R. Leuze and Daniel A. Reed, editors, Proceedings of the 6th SIAM Conference on Parallel Processing for Scientific Computing, pages 913--922, Norfolk, VI, March 1993. SIAM Press.

....establishments, resource management of these machines becomes a more pressing issue. Job scheduling is an important aspect of this work, and much research on the topic of job scheduling for parallel computers has been done in recent years. Some have focused on distributed memory machines [7, 8, 9, 11, 14], others on more general systems [3, 6, 10, 15] Many different strategies (static vs. dynamic, time sharing vs. space sharing, etc. have been investigated and compared. All these studies make underlying assumptions about the workload in order to quantify their results. Some of these studies use ....

V.K. Naik, S.K. Setia and M.S. Squillante. "Scheduling of Large Scientific applications on distributed Memory Multiprocessor Systems". In Sixth SIAM Conference on Parallel Processing for Scientific Computing, 1993.

....use. It is safe to speculate that this will also be the primary use of parallel workstations. Other systems attempt to combine interactive and batch usage modes. For example, it is possible to partition the machine statically into a large batch partition and a smaller interactive partition [400, 422, 606]. Alternatively, it is possible to initially allocate all the resources to batch processing, and preempt PEs and memory in favor of interactive jobs as needed. This leads to the notion of adaptive resource allocation for the batch jobs, using the same techniques as for interactive jobs (e.g. time ....

....that run on systems that change the PE allocation at runtime must be able to adjust to whatever number of PEs the system gives them. We shall classify the partitioning mechanisms into four types, based on how the operating system behaves and the requirements this places on applications [422, 489]: fixed, variable, adaptive, and dynamic. These are summarized in Table 1 and discussed in Subsections 3.1 through 3.4, respectively. As indicated in the table, the progress towards dynamic partitioning is motivated by the desire to eliminate fragmentation and improve resource utilization. ....

[Article contains additional citation context not shown here]

V. K. Naik, S. K. Setia, and M. S. Squillante, "Scheduling of large scientific applications on distributed memory multiprocessor systems". In 6th SIAM Conf. Parallel Processing for Scientific Computing, vol. II, pp. 913--922, Mar 1993.

....that among dynamic policies that do not use job demand information, policies that allocate processing power approximately equally among executing jobs (i. e, EQ policies) have high performance when the coefficient of variation, C D , in cumulative job processing requirement is greater than one [LV90, ST91, NSS93a, NSS93b, MVZ93, MV93b]. Also, several studies have shown that particular RTC policies have poorer performance than particular dynamic policies [ZM90, MEB91, ST91, NSS93b] As a third example, there is consensus in the literature that RTC policies have higher performance if they allocate fewer processors to a job as ....

.... RTC policies have poorer performance than particular dynamic policies [ZM90, MEB91, ST91, NSS93b] As a third example, there is consensus in the literature that RTC policies have higher performance if they allocate fewer processors to a job as (instantaneous or average) system load increases [Se89, GST91, NSS93a, NSS93b, ST93]. The data in [ZM90, MEB91, RSDS93, SRDS93] also support this conclusion. Questions that remain unresolved in the literature include: 1) Should RTC policies use application execution rate characteristics, such as average parallelism (avg) or processor working set (pws) defined in section 2) in ....

[Article contains additional citation context not shown here]

V. Naik, S. Setia, and M. Squillante. Scheduling of Large Scientific Applications on Distributed Memory Multiprocessor Systems. Proc. of the 6th SIAM Conference on Parallel Processing for Scientific Computation, 1993.

....partitioning schemes is complicated by the fact that three players are involved: the architecture, the operating system, and the applications. We shall classify the partitioning mechanisms into four types, based on how the operating system behaves and the requirements this places on applications [260, 295]: fixed, variable, adaptive, and dynamic. These are summarized in Table 1 and discussed in Subsections 3.1 through 3.4, respectively. As indicated in the table, the progress towards dynamic partitioning is motivated by the desire to eliminate fragmentation and improve resource utilization. ....

....the system administrator, typically for reasons related to access control. This allows certain parts of the machine to be dedicated to certain groups of users, possibly according to their investment in procuring the machine. Alternatively, the partitions can be designated for different job classes [243, 260]. For example, many commercial parallel 10 type operating system application runtime advantages disadvantages fixed predefined partitions parallelism hardcoded or a parameter simple, preserves locality internal fragmentation, limited multiprogramming, arbitrary queueing variable ....

[Article contains additional citation context not shown here]

V. K. Naik, S. K. Setia, and M. S. Squillante, "Scheduling of large scientific applications on distributed memory multiprocessor systems". In 6th SIAM Conf. Parallel Processing for Scientific Computing, vol. II, pp. 913--922, Mar 1993. 90

....Foundation under grants CCR 9024144 and CDA 9024618. 1 1 Introduction The increasing use of parallel processor systems has led to the development of a number of multiprogrammed processor allocation policies. Many studies have compared the performance of specific processor allocation policies [7, 13, 14, 15, 16, 20, 21, 22, 25, 33, 34, 35, 37, 39, 41], which has led to a diverse set of results concerning relative policy performance over numerous specific regions of the workload parameter space. 1 For example, ffl The Adaptive Static Partitioning (ASP) policy has been shown to have higher performance than several static allocation policies, ....

.... and highly correlated workloads when CD is moderate to high [14] ffl The spatial equipartitioning (EQS) policy is also shown to have high performance for specific measurement workloads [39, 20, 8] and a particular workload that consists of a mixture of application types and has high CD [21, 22]. These results show particular policies to perform well over narrow regions of the parameter space but it is not clear whether or how the various results generalize. In particular, it is not clear which workload parameters determine relative policy performance or how the policies compare for say ....

[Article contains additional citation context not shown here]

V. Naik, S. Setia, and M. Squillante. Scheduling of Large Scientific Applications on Distributed Memory Multiprocessor Systems. Proceedings of the 6th SIAM Conference on Parallel Processing for Scientific Computation. IBM Research Report RC 18621, T. J. Watson Research Center, Yorktown Heights, Jan. 1993.

....scheduling scheme for small scale uniform memory access (UMA) multiprocessors, provided applications are coded in a style that can tolerate dynamic changes in the number of processors at runtime. Under certain circumstances, its use can be extended to large distributed memory machines as well [26, 24]. A number of papers have evaluated the performance implications of coscheduling and gang scheduling, and compared them with dynamic partitioning and other scheduling policies for multiprogrammed multiprocessors [10, 28, 22, 16, 35] The Mach scheduler also allows processors to be allocated to ....

V. K. Naik, S. K. Setia, and M. S. Squillante, "Scheduling of large scientific applications on distributed memory multiprocessor systems". In 6th SIAM Conf. Parallel Processing for Scientific Computing, vol. II, pp. 913--922, Mar 1993.

....and CDA 9024618. 1 Introduction The algorithm for scheduling jobs on the processors of a multiprogrammed parallel computer can have a significant impact on system performance. Parallel processor scheduling disciplines have been studied using system measurement [10, 23, 44, 46] simulation [9, 17, 18, 24, 50], and analytic modeling methods [8, 16, 19, 25, 27, 28, 34, 35, 42] While these studies have yielded various specific insights, the general performance characteristics of parallel scheduling policies still remains poorly understood. Questions such as which policies dominate over various regions ....

....a job is never allocated more processors than its available parallelism. An example of such a policy is the default CM 5 scheduler for jobs that fit in the memory of a single partition. Previous studies have shown that variants of the EQ policy have high performance under a variety of workloads [44, 17, 23, 10, 16, 35, 24]. Finally, we examine the PSAPF policy proposed in [18] because of its potential for high performance for workloads where job processing time is correlated with parallelism [18, 16] Furthermore, this policy allows us to illustrate an interesting aspect of the interpolation approximation approach. ....

[Article contains additional citation context not shown here]

V. Naik, S. Setia and M. Squillante. Scheduling of Large Scientific Applications on Distributed Memory Multiprocessor Systems. Research Report RC 18621, IBM T. J. Watson Research Center, Yorktown Heights, Jan. 1993. Proc. of the 6th SIAM Conf. on Parallel Processing for Scientific Computation.

....the end of an interval, all processors are reallocated to jobs in the next shelf. They prove their self scheduling algorithm finds the minimal length schedule within the class of shelf algorithms. There has been limited experience with dynamic allocation in distributed memory systems. Setia et al. [29] compare a dynamic allocation policy to static partitioning and adaptive partitioning for varying workloads. Under their dynamic allocation policy, partition sizes are changed upon arrival and departure to accommodate newly arriving jobs or to utilize processors freed by departing jobs. They ....

....than do the other policies. Their policies do not consider adjacency of multiple processors allocated to a job, and thus ignore the potential performance impact of interconnection network contention among multiple applications competing for the same interconnection network links. Several studies [30, 29, 38] explicitly demonstrate the necessity of reducing the allocation of processors to jobs as the system load increases. Peris et al. 33] examine the effects of increased paging overhead accompanying a reduced allocation on job performance. Their results illustrate the necessity to consider the ....

V. Naik, S. Setia, and M. Squillante. Scheduling of large scientific applications on distributed memory multiprocessor systems. In Proceedings of the 6th SIAM Conference on Parallel Processing for Scientific Computation, pages 913--922, March 1993.

....and CDA 9024618. 1 Introduction Dynamic equiallocation policies are a class of parallel processor scheduling policies that attempt to allocate processing power equally to all jobs, subject to the constraint that no job is allocated more processors than its available parallelism. Several studies [30, 11, 10, 18, 7, 27, 19, 20, 16] of multiprogrammed parallel systems have observed that dynamic equiallocation policies have high performance for a variety of workloads, where the performance metric is mean response time. However, the performance characteristics of equiallocation policies are not yet thoroughly understood. ....

V. Naik, S. Setia, and M. Squillante. Scheduling of Large Scientific Applications on Distributed Memory Multiprocessor Systems. Proceedings of the 6th SIAM Conference on Parallel Processing for Scientific Computation. IBM Research Report RC 18621, T. J. Watson Research Center, Yorktown Heights, Jan. 1993.

....employed in a number of commercial systems. This is due in part to its low system overhead and its simplicity from both the system and application viewpoints. The static scheduling approach, however, can lead to relatively low system throughputs and resource utilizations under nonuniform workloads [34, 21, 22, 25, 35], as is common in scientific engineering computing environments [6] Adaptive partitioning policies, where the number of processors allocated to a job is determined when jobs arrive and depart based on the current system state, have also been considered in a number of research studies [14, 42, 8, ....

.... 22, 25, 35] as is common in scientific engineering computing environments [6] Adaptive partitioning policies, where the number of processors allocated to a job is determined when jobs arrive and depart based on the current system state, have also been considered in a number of research studies [14, 42, 8, 21, 22, 33, 3, 25]. This approach tends to outperform its static counterparts by adapting partition sizes to the current load. However, the performance benefits of adaptive partitioning can be limited due to its inability to adjust scheduling decisions in response to subsequent workload changes. These potential ....

[Article contains additional citation context not shown here]

V. K. Naik, S. K. Setia, and M. S. Squillante. Scheduling of large scientific applications on distributed memory multiprocessor systems. In Proc. SIAM Conf. Par. Proc. Sci. Comp., 913--922, 1993.

....employed in a number of commercial systems. This is due in part to its low system overhead and its simplicity from both the system and application viewpoints. The static scheduling approach, however, can lead to relatively low system throughputs and resource utilizations under nonuniform workloads [33, 26, 27, 30, 34], which can be common in scientific and engineering computing environments. Adaptive partitioning policies, where the number of nodes allocated to a job is determined when jobs enter and leave based on the current system state, have also been considered in a number of research studies [21, 38, 16, ....

....27, 30, 34] which can be common in scientific and engineering computing environments. Adaptive partitioning policies, where the number of nodes allocated to a job is determined when jobs enter and leave based on the current system state, have also been considered in a number of research studies [21, 38, 16, 26, 27, 32, 7, 30]. This approach tends to outperform its static counterparts by adapting partition sizes to the current load. On the other hand, the performance benefits of adaptive partitioning can be limited due to its inability to adjust scheduling decisions in response to subsequent workload changes. These ....

[Article contains additional citation context not shown here]

V. K. Naik, S. K. Setia, and M. S. Squillante. Scheduling of large scientific applications on distributed memory multiprocessor systems. In Proceedings Sixth SIAM Conference on Parallel Processing for Scientific Computing, pages 913--922, March 1993.

....in a number of commercial systems. This is due in part to its low system overhead and its simplicity from both the system and application viewpoints. The static scheduling approach, however, can lead to relatively low system throughputs and resource utilizations under nonuniform workloads [35, 26, 27, 31, 36, 28], which can be common in scientific engineering computing environments. Adaptive partitioning policies, where the number of nodes allocated to a job is determined at job arrival instants based on the current system state, have also been considered in a number of research studies [21, 40, 15, 26, ....

....26, 27, 31, 36, 28] which can be common in scientific engineering computing environments. Adaptive partitioning policies, where the number of nodes allocated to a job is determined at job arrival instants based on the current system state, have also been considered in a number of research studies [21, 40, 15, 26, 27, 34, 7, 31, 28]. This approach tends to outperform its static counterparts by adapting partition sizes to the current load. On the other hand, the performance benefits of adaptive partitioning can be limited due to its inability to adjust scheduling decisions in response to subsequent workload changes. These ....

[Article contains additional citation context not shown here]

V. K. Naik, S. K. Setia, and M. S. Squillante. Scheduling of large scientific applications on distributed memory multiprocessor systems. In Proceedings Sixth SIAM Conference on Parallel Processing for Scientific Computing, pages 913--922, March 1993.

No context found.

V. K. Naik, S. K. Setia, and M. S. Squillante. Scheduling of large scientific applications on distributed memory multiprocessor systems. In Proceedings Sixth SIAM Conference on Parallel Processing for Scientific Computing, pages 913--922, March 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC