The Experts below are selected from a list of 41973 Experts worldwide ranked by ideXlab platform
Stijn Eyerman - One of the best experts on this subject based on the ideXlab platform.
-
HPCA - Symbiotic Job Scheduling on the IBM POWER8
2016 IEEE International Symposium on High Performance Computer Architecture (HPCA), 2016Co-Authors: Josue Feliu, Stijn Eyerman, Julio Sahuquillo, Salvador PetitAbstract:Simultaneous multithreading (SMT) processors share most of the microarchitectural core components among the co-running applications. The competition for shared resources causes performance interference between applications. Therefore, the performance benefits of SMT processors heavily depend on the complementarity of the co-running applications. Symbiotic Job Scheduling, i.e., Scheduling applications that co-run well together on a core, can have a considerable impact on the performance of a processor with SMT cores. Prior work uses sampling or novel hardware support to perform symbiotic Job Scheduling, which has either a non-negligible overhead or is impossible to use on existing hardware. This paper proposes a symbiotic Job scheduler for the IBM POWER8 processor. We leverage the existing cycle accounting mechanism to predict symbiosis between applications, and use that information at run-time to decide which applications should run on the same core or on separate cores. We implement the scheduler in the Linux operating system and evaluate it on an IBM POWER8 server running multiprogrammed workloads. The symbiotic Job scheduler significantly improves performance compared to both an agnostic random scheduler and the default Linux scheduler. With respect to Linux, it achieves an average speedup by 8.8% for workloads comprising 12 applications, and by 4.7% on average across all evaluated workloads.
-
Revisiting Symbiotic Job Scheduling
2015Co-Authors: Stijn Eyerman, Pierre Michaud, Wouter RogiestAbstract:—Symbiotic Job Scheduling exploits the fact that in a system with shared resources, the performance of Jobs is impacted by the behavior of other co-running Jobs. By coScheduling combinations of Jobs that have low interference, the performance of a system can be increased. In this paper, we investigate the impact of using symbiotic Job Scheduling for increasing throughput. We find that even for a theoretically optimal scheduler, this impact is very low, despite the substantial sensitivity of per Job performance to which other Jobs are coscheduled: for example, our experiments on a 4-thread SMT processor show that, on average, the Job IPC varies by 37% depending on coscheduled Jobs, the per-coschedule throughput varies by 69%, and yet the average throughput gain brought by optimal symbiotic Scheduling is only 3%. This small margin of improvement can be explained by the observation that all the Jobs need to be eventually executed, restricting the Job combinations a symbiotic Job scheduler can select to optimize throughput. We explain why previous work reported a substantial gain from symbiotic Job Scheduling, and we find that (only) reporting turnaround time can lead to misleading conclusions. Furthermore , we show how the impact of Scheduling can be evaluated in microarchitectural studies, without having to implement a scheduler.
-
ISPASS - Revisiting symbiotic Job Scheduling
2015 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), 2015Co-Authors: Stijn Eyerman, Pierre Michaud, Wouter RogiestAbstract:Symbiotic Job Scheduling exploits the fact that in a system with shared resources, the performance of Jobs is impacted by the behavior of other co-running Jobs. By coScheduling combinations of Jobs that have low interference, the performance of a system can be increased. In this paper, we investigate the impact of using symbiotic Job Scheduling for increasing throughput. We find that even for a theoretically optimal scheduler, this impact is very low, despite the substantial sensitivity of per Job performance to which other Jobs are coscheduled: for example, our experiments on a 4-thread SMT processor show that, on average, the Job IPC varies by 37% depending on coscheduled Jobs, the per-coschedule throughput varies by 69%, and yet the average throughput gain brought by optimal symbiotic Scheduling is only 3%. This small margin of improvement can be explained by the observation that all the Jobs need to be eventually executed, restricting the Job combinations a symbiotic Job scheduler can select to optimize throughput. We explain why previous work reported a substantial gain from symbiotic Job Scheduling, and we find that (only) reporting turnaround time can lead to misleading conclusions. Furthermore, we show how the impact of Scheduling can be evaluated in microarchitectural studies, without having to implement a scheduler.
Ruay-shiung Chang - One of the best experts on this subject based on the ideXlab platform.
-
an adaptive scoring Job Scheduling algorithm for grid computing
Information Sciences, 2012Co-Authors: Ruay-shiung ChangAbstract:When human culture advances, current problems in science and engineering become more complicated and need more computing power to tackle and analyze. A supercomputer is not the only choice for solving complex problems any more as a result of the speed-up of personal computers and networks. Grid technology, which connects a number of personal computer clusters with high speed networks, can achieve the same computing power as a supercomputer does, also with a lower cost. However, grid is a heterogeneous system. Scheduling independent tasks on it is more complicated. In order to utilize the power of grid completely, we need an efficient Job Scheduling algorithm to assign Jobs to resources in a grid. In this paper, we propose an Adaptive Scoring Job Scheduling algorithm (ASJS) for the grid environment. Compared to other methods, it can decrease the completion time of submitted Jobs, which may compose of computing-intensive Jobs and data-intensive Jobs.
-
Improving Job Scheduling algorithms in a grid environment
Future Generation Computer Systems, 2011Co-Authors: Yun-han Lee, Seiven Leu, Ruay-shiung ChangAbstract:Due to the advances in human civilization, problems in science and engineering are becoming more complicated than ever before. To solve these complicated problems, grid computing becomes a popular tool. A grid environment collects, integrates, and uses heterogeneous or homogeneous resources scattered around the globe by a high-speed network. A grid environment can be classified into two types: computing grids and data grids. This paper mainly focuses on computing grids. In computing grid, Job Scheduling is a very important task. A good Scheduling algorithm can assign Jobs to resources efficiently and can balance the system load. In this paper, we propose a hierarchical framework and a Job Scheduling algorithm called Hierarchical Load Balanced Algorithm (HLBA) for Grid environment. In our algorithm, we use the system load as a parameter in determining a balance threshold. And the scheduler adapts the balance threshold dynamically when the system load changes. The main contributions of this paper are twofold. First, the Scheduling algorithm balances the system load with an adaptive threshold and second, it minimizes the makespan of Jobs. Experimental results show that the performance of HLBA is better than those of other algorithms.
-
an ant algorithm for balanced Job Scheduling in grids
Future Generation Computer Systems, 2009Co-Authors: Ruay-shiung Chang, Jihsheng Chang, Posheng LinAbstract:Grid computing utilizes the distributed heterogeneous resources in order to support complicated computing problems. Grid can be classified into two types: computing grid and data grid. Job Scheduling in computing grid is a very important problem. To utilize grids efficiently, we need a good Job Scheduling algorithm to assign Jobs to resources in grids. In the natural environment, the ants have a tremendous ability to team up to find an optimal path to food resources. An ant algorithm simulates the behavior of ants. In this paper, we propose a Balanced Ant Colony Optimization (BACO) algorithm for Job Scheduling in the Grid environment. The main contributions of our work are to balance the entire system load while trying to minimize the makespan of a given set of Jobs. Compared with the other Job Scheduling algorithms, BACO can outperform them according to the experimental results.
Salvador Petit - One of the best experts on this subject based on the ideXlab platform.
-
HPCA - Symbiotic Job Scheduling on the IBM POWER8
2016 IEEE International Symposium on High Performance Computer Architecture (HPCA), 2016Co-Authors: Josue Feliu, Stijn Eyerman, Julio Sahuquillo, Salvador PetitAbstract:Simultaneous multithreading (SMT) processors share most of the microarchitectural core components among the co-running applications. The competition for shared resources causes performance interference between applications. Therefore, the performance benefits of SMT processors heavily depend on the complementarity of the co-running applications. Symbiotic Job Scheduling, i.e., Scheduling applications that co-run well together on a core, can have a considerable impact on the performance of a processor with SMT cores. Prior work uses sampling or novel hardware support to perform symbiotic Job Scheduling, which has either a non-negligible overhead or is impossible to use on existing hardware. This paper proposes a symbiotic Job scheduler for the IBM POWER8 processor. We leverage the existing cycle accounting mechanism to predict symbiosis between applications, and use that information at run-time to decide which applications should run on the same core or on separate cores. We implement the scheduler in the Linux operating system and evaluate it on an IBM POWER8 server running multiprogrammed workloads. The symbiotic Job scheduler significantly improves performance compared to both an agnostic random scheduler and the default Linux scheduler. With respect to Linux, it achieves an average speedup by 8.8% for workloads comprising 12 applications, and by 4.7% on average across all evaluated workloads.
Josue Feliu - One of the best experts on this subject based on the ideXlab platform.
-
HPCA - Symbiotic Job Scheduling on the IBM POWER8
2016 IEEE International Symposium on High Performance Computer Architecture (HPCA), 2016Co-Authors: Josue Feliu, Stijn Eyerman, Julio Sahuquillo, Salvador PetitAbstract:Simultaneous multithreading (SMT) processors share most of the microarchitectural core components among the co-running applications. The competition for shared resources causes performance interference between applications. Therefore, the performance benefits of SMT processors heavily depend on the complementarity of the co-running applications. Symbiotic Job Scheduling, i.e., Scheduling applications that co-run well together on a core, can have a considerable impact on the performance of a processor with SMT cores. Prior work uses sampling or novel hardware support to perform symbiotic Job Scheduling, which has either a non-negligible overhead or is impossible to use on existing hardware. This paper proposes a symbiotic Job scheduler for the IBM POWER8 processor. We leverage the existing cycle accounting mechanism to predict symbiosis between applications, and use that information at run-time to decide which applications should run on the same core or on separate cores. We implement the scheduler in the Linux operating system and evaluate it on an IBM POWER8 server running multiprogrammed workloads. The symbiotic Job scheduler significantly improves performance compared to both an agnostic random scheduler and the default Linux scheduler. With respect to Linux, it achieves an average speedup by 8.8% for workloads comprising 12 applications, and by 4.7% on average across all evaluated workloads.
Wouter Rogiest - One of the best experts on this subject based on the ideXlab platform.
-
Revisiting Symbiotic Job Scheduling
2015Co-Authors: Stijn Eyerman, Pierre Michaud, Wouter RogiestAbstract:—Symbiotic Job Scheduling exploits the fact that in a system with shared resources, the performance of Jobs is impacted by the behavior of other co-running Jobs. By coScheduling combinations of Jobs that have low interference, the performance of a system can be increased. In this paper, we investigate the impact of using symbiotic Job Scheduling for increasing throughput. We find that even for a theoretically optimal scheduler, this impact is very low, despite the substantial sensitivity of per Job performance to which other Jobs are coscheduled: for example, our experiments on a 4-thread SMT processor show that, on average, the Job IPC varies by 37% depending on coscheduled Jobs, the per-coschedule throughput varies by 69%, and yet the average throughput gain brought by optimal symbiotic Scheduling is only 3%. This small margin of improvement can be explained by the observation that all the Jobs need to be eventually executed, restricting the Job combinations a symbiotic Job scheduler can select to optimize throughput. We explain why previous work reported a substantial gain from symbiotic Job Scheduling, and we find that (only) reporting turnaround time can lead to misleading conclusions. Furthermore , we show how the impact of Scheduling can be evaluated in microarchitectural studies, without having to implement a scheduler.
-
ISPASS - Revisiting symbiotic Job Scheduling
2015 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), 2015Co-Authors: Stijn Eyerman, Pierre Michaud, Wouter RogiestAbstract:Symbiotic Job Scheduling exploits the fact that in a system with shared resources, the performance of Jobs is impacted by the behavior of other co-running Jobs. By coScheduling combinations of Jobs that have low interference, the performance of a system can be increased. In this paper, we investigate the impact of using symbiotic Job Scheduling for increasing throughput. We find that even for a theoretically optimal scheduler, this impact is very low, despite the substantial sensitivity of per Job performance to which other Jobs are coscheduled: for example, our experiments on a 4-thread SMT processor show that, on average, the Job IPC varies by 37% depending on coscheduled Jobs, the per-coschedule throughput varies by 69%, and yet the average throughput gain brought by optimal symbiotic Scheduling is only 3%. This small margin of improvement can be explained by the observation that all the Jobs need to be eventually executed, restricting the Job combinations a symbiotic Job scheduler can select to optimize throughput. We explain why previous work reported a substantial gain from symbiotic Job Scheduling, and we find that (only) reporting turnaround time can lead to misleading conclusions. Furthermore, we show how the impact of Scheduling can be evaluated in microarchitectural studies, without having to implement a scheduler.
