Heterogeneity-Aware Resource Allocation in HPC Systems
In their march towards exascale performance, HPC systems are becoming increasingly more heterogeneous in an effort to keep power consumption at bay. Exploiting accelerators such as GPUs and MICs together with traditional processors to their fullest requires heterogeneous HPC systems to employ intelligent job dispatchers that go beyond the capabilities of those that have been developed for homogeneous systems. In this paper, we propose three new heterogeneity-aware resource allocation algorithms suitable for building job dispatchers for any HPC system. We use real workload traces extracted from the Eurora HPC system to analyze the performance of our allocators when they are coupled with different schedulers. Our experimental results show that significant improvements can be obtained in job response times and system throughput over solutions developed for homogeneous systems. Our study also helps to characterize the operating conditions in which heterogeneity-aware resource allocation becomes crucial for heterogeneous HPC systems.
We thank Dr. A. Bartolini, Prof. L. Benini, Prof. M. Milano, Dr. M. Lombardi and the SCAI group at Cineca for providing access to the Eurora data. We also thank the IT Center of the University of Pisa (Centro Interdipartimentale di Servizi e Ricerca) for providing access to computing resources for simulations. A. Netti has been supported by a research fellowship from the Oprecomp-Open Transprecision Computing project. C. Galleguillos has been supported by Postgraduate Grant PUCV 2017. A. Sîrbu has been partially funded by the EU project SoBigData Research Infrastructure—Big Data and Social Mining Ecosystem (grant agreement 654024).
- 1.Ashby, S., Beckman, P., Chen, J., Colella, P., Collins, B., Crawford, D., et al.: The opportunities and challenges of exascale computing. Summary Report of the Advanced Scientific Computing Advisory Committee (ASCAC) Subcommittee, pp. 1–77 (2010)Google Scholar
- 3.Bhattacharya, S., Tsai, W.: Lookahead processor allocation in mesh-connected massively parallel multicomputer. In: Proceedings of IPPS 1994, pp. 868–875. IEEE (1994)Google Scholar
- 7.Cavazzoni, C.: Eurora: a European architecture toward exascale. In: Future HPC Systems: The Challenges of Power-Constrained Performance. ACM (2012)Google Scholar
- 11.Galleguillos, C., Sîrbu, A., Kiziltan, Z., Babaoglu, O., Borghesi, A., Bridi, T.: Data-driven job dispatching in HPC systems. In: Nicosia, G., Pardalos, P., Giuffrida, G., Umeton, R. (eds.) MOD 2017. LNCS, vol. 10710, pp. 449–461. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-72926-8_37CrossRefGoogle Scholar
- 13.Guim, F., Rodero, I., Corbalan, J., Parashar, M.: Enabling GPU and many-core systems in heterogeneous HPC environments using memory considerations. In: Proceedings of HPCC 2010, pp. 146–155. IEEE (2010)Google Scholar
- 17.Reuther, A., Byun, C., Arcand, W., Bestor, D., Bergeron, B., Hubbell, M., et al.: Scalable system scheduling for HPC and big data. arXiv:1705.03102 (2017)
- 20.Villa, O., Johnson, D.R., Oconnor, M., Bolotin, E., Nellans, D., Luitjens, J., et al.: Scaling the power wall: a path to exascale. In: Proceedings of SC 2014, pp. 830–841. IEEE (2014)Google Scholar
- 21.Wong, A.K.L., Goscinski, A.M.: Evaluating the EASY-backfill job scheduling of static workloads on clusters. In: Proceedings of CLUSTER 2007, pp. 64–73. IEEE (2007)Google Scholar
- 23.Zeldes, Y., Feitelson, D.G.: On-line fair allocations based on bottlenecks and global priorities. In: Proceedings of ICPE 2013, pp. 229–240. ACM (2013)Google Scholar