Automating Job Monitoring System for an Ecosystem of High Performance Computing

ABSTRACT

Many countries have founded national high performance computing center aiming to provide computational resources to their scientists upon requests. The resources provided are not efficient because the job requests are not relative to the real use leading to unnecessary resource consumption. In this paper, we present a method to monitor and manage High Performance Computing (HPC) resources more efficiently. Usually, the HPC resources are managed by a Portable Batch System (PBS) as the Job Management System (JMS) for effective job scheduling and resource allocation. However, the HPC resources often engage in inefficient job requests. For instance, a job request may have for four processors running per node for two hours, but the actual usage engages four processors per node for one hour. Hence, the HPC resources lose an hour of productivity. As a consequence, the queues for job execution are longer. The automated job monitoring system proposed in this paper would scan all the jobs on every HPC Node and compare the job requests conditions with preset criteria. If the conditions meet the criteria, then the inefficient jobs are forced to cancel from the HPC queue. The results show that more HPC resources are available for executing other jobs in the queue, leading to saved resources in the HPC environment and Stabilization of HPC hardware, promoting an HPC infrastructure ecosystem.

References

1. Aida, K. 2000. Effect of job size characteristics on job scheduling performance. Lecture notes in computer science. 1911, (2000), 1--17.
2. CERN | Accelerating science: https://home.cern/. Accessed: 2017-06-24.
3. Downey, A.B. 1997. A parallel workload model and its implications for processor allocation. High Performance Distributed Computing, 1997. Proceedings. The Sixth IEEE International Symposium on (1997), 112--123.
4. Feitelson, D.G. 1996. Packing schemes for gang scheduling. Job Scheduling Strategies for Parallel Processing: IPPS '96 Workshop Honolulu, Hawaii, April 16, 1996 Proceedings. D.G. Feitelson and L. Rudolph, eds. Springer Berlin Heidelberg. 89--110.
5. Home - National e-Science Infrastructure Consortium: http://www.e-science.in.th/infra/. Accessed: 2016-07-06.
6. Hovestadt, M. et al. 2003. Scheduling in HPC Resource Management Systems: Queuing vs. Planning. Job Scheduling Strategies for Parallel Processing: 9th International Workshop, JSSPP 2003, Seattle, WA, USA, June 24, 2003. Revised Paper. D. Feitelson et al., eds. Springer Berlin Heidelberg. 1--20.
7. Job Management Systems: http://www.cro-ngi.hr/en/technologies/cluster-technologies/job-management-systems/. Accessed: 2017-03-07.
8. Lifka, D.A. 1995. The anl/ibm sp scheduling system. Workshop on Job Scheduling Strategies for Parallel Processing (1995), 295--303.
9. Lo, V. et al. 1998. A comparative study of real workload traces and synthetic workload models for parallel job scheduling. Workshop on Job Scheduling Strategies for Parallel Processing (1998), 25--46.
10. Skovira, J. et al. 1996. The EASY---LoadLeveler API Project. Job Scheduling Strategies for Parallel Processing (1996), 41--47.
11. Subhlok, J. et al. 1996. Impact of job mix on optimizations for space sharing schedulers. Supercomputing, 1996. Proceedings of the 1996 ACM/IEEE Conference on (1996), 54--54.
12. V. Subramani et al. 2002. Distributed job scheduling on computational Grids using multiple simultaneous requests. Proceedings 11th IEEE International Symposium on High Performance Distributed Computing (2002), 359--366.
13. Welcome to Python.org: https://www.python.org/. Accessed: 2017-06-26.
14. What is a Bash Script? - Bash Scripting Tutorial: http://ryanstutorials.net/bash-scripting-tutorial/bash-script.php. Accessed: 2017-06-26.
15. Yan, Y. and Chapman, B. 2008. Comparative Study of Distributed Resource Management Systems--SGE, LSF, PBS Pro, and LoadLeveler. Technical Report-Citeseerx. (2008).

• General and reference

  • Cross-computing tools and techniques

    • Performance