Learning budget assignment policies for autoscaling scientific workflows in the cloud

  • Yisel Garí
  • David A. MongeEmail author
  • Cristian Mateos
  • Carlos García Garino


Autoscalers exploit cloud-computing elasticity to cope with the dynamic computational demands of scientific workflows. Autoscalers constantly acquire or terminate virtual machines (VMs) on-the-fly to execute workflows minimizing makespan and economic cost. One key problem of workflow autoscaling under budget constraints (i.e. with a maximum limit in cost) is determining the right proportion between: (a) expensive but reliable VMs called on-demand instances, and (b) cheaper but subject-to-failure VMs called spot instances. Spot instances can potentially provide huge parallelism possibilities at low costs but they must be used wisely as they can fail unexpectedly hindering makespan. Given the unpredictability of failures and the inherent performance variability of clouds, designing a policy for assigning the budget for each kind of instance is not a trivial task. For such reason we formalize the described problem as a Markov decision process that allows us learning near-optimal policies from the experience of other baseline policies. Experiments over four well-known scientific workflows, demonstrate that learned policies outperform the baseline policies considering the aggregated relative percentage difference of makespan and execution cost. These promising results encourage the future study of new strategies aiming to find optimal budget policies applied to the execution of workflows in the cloud.


Scientific workflows Cloud computing Autoscaling Spot instances Markov decision process 



This research is supported by the ANPCyT Projects No. PICT-2012-2731 and PICT-2014-1430, and by the UNCuyo Project No. SeCTyP-M041. The first author acknowledges her PhD fellowship granted by the CONICET.


  1. 1.
    Buyya, R., Yeo, C., Venugopal, S., Broberg, J., Brandic, I.: Cloud computing and emerging IT platforms: Vision, hype, and reality for delivering computing as the 5th utility. Future Gen. Comput. Syst. 25(6), 599–616 (2009)CrossRefGoogle Scholar
  2. 2.
    Mao, M., Humphrey, M.: Scaling and scheduling to maximize application performance within budget constraints in cloud workflows. In: IEEE 27th International Symposium on Parallel & Distributed Processing (IPDPS), 2013, pp. 67–78. IEEE (2013)Google Scholar
  3. 3.
    Monge, D.A., Garí, Y., Mateos, C., Garino, C.G.: Autoscaling scientific workflows on the cloud by combining on-demand and spot instances. Int. J. Comput. Syst. Sci. Eng. 32(4), 291–306 (2017). (Special Issue on Elastic Data Management in Cloud Systems)Google Scholar
  4. 4.
    Expósito, R.R., Taboada, G.L., Ramos, S., Touriño, J., Doallo, R.: Performance analysis of HPC applications in the cloud. Future Gen. Comput. Syst. 29(1), 218–229 (2013)CrossRefGoogle Scholar
  5. 5.
    Ben-Yehuda, O.A., Ben-Yehuda, M., Schuster, A., Tsafrir, D.: Deconstructing amazon EC2 spot instance pricing. ACM Trans. Econ. Comput. 1(3), 16:1–16:20 (2013)Google Scholar
  6. 6.
    Ghobaei-Arani, M., Jabbehdari, S., Pourmina, M.A.: An autonomic resource provisioning approach for service-based cloud applications: a hybrid approach. Future Gen. Comput. Syst. 78, 191–210 (2018)CrossRefGoogle Scholar
  7. 7.
    Benifa, J.V.B., Dejey, D.: RLPAS: reinforcement learning-based proactive auto-scaler for resource provisioning in cloud environment. Mob. Netw. Appl. 2018, 1–16 (2018)Google Scholar
  8. 8.
    Soualhia, M., Khomh, F., Tahar, S.: A dynamic and failure-aware task scheduling framework for hadoop. IEEE Trans. Cloud Comput. (early access) 1–16 (2018)Google Scholar
  9. 9.
    Garí, Y., Monge, D.A., Mateos, C., Garino, C.G.: Markov Decision Process to Dynamically Adapt Spots Instances Ratio on the Autoscaling of Scientific Workflows in the Cloud, pp. 353–369. Springer, Cham (2018)Google Scholar
  10. 10.
    Juve, G., Chervenak, A., Deelman, E., Bharathi, S., Mehta, G., Vahi, K.: Characterizing and profiling scientific workflows. Future Gen. Comput. Syst. 29(3), 682–692 (2013)CrossRefGoogle Scholar
  11. 11.
    Monge, D.A., Holec, M., Železnỳ, F., Garino, C.G.: Ensemble learning of runtime prediction models for gene-expression analysis workflows. Clust. Comput. 18(4), 1317–1329 (2015)CrossRefGoogle Scholar
  12. 12.
    Turchenko, V., Shultz, V., Turchenko, I., Wallace, R.M., Sheikhalishahi, M., Vazquez-Poletti, J.L., Lucio, G.: Spot price prediction for cloud computing using neural networks. Int. J. Comput. 12(4), 348–359 (2013)Google Scholar
  13. 13.
    Shaojie, T., Jing, Y., Li, X.Y.: Towards optimal bidding strategy for Amazon EC2 cloud spot instance. In: Proceedings 2012 IEEE 5th International Conference on Cloud Computing, CLOUD 2012, pp. 91–98 (2012)Google Scholar
  14. 14.
    Bellman, R.: Dynamic Programming. Princeton University Press, New Jersey (1957)zbMATHGoogle Scholar
  15. 15.
    Van Otterlo, M.: The Logic of Adaptive Behavior. In: Frontiers in Artificial Intelligence and Applications, vol. 192. IOS Press, Amsterdam (2009)Google Scholar
  16. 16.
    Enda, B., Enda, H., Jim, D.: A learning architecture for scheduling workflow applications in the cloud. Proceedings 9th IEEE European Conference on Web Services, ECOWS 2011, pp. 83–90 (2011)Google Scholar
  17. 17.
    Enda, B., Enda, H., Jim, D.: Applying reinforcement learning towards automating resource allocation and application scalability in the cloud. Concurr. Comput. Pract. Exp. 24(13), 1397–1420 (2012)CrossRefGoogle Scholar
  18. 18.
    Philip, M., Ewa, D., Li, Z., Robert, G., Gaurang, M., Nitin, G., John, M., Carl, K., Scott, C., David, O., Hunter, F., Vipin, G., Yifeng, C., Karan, V., Thomas, J., Edward, F.: SCEC CyberShake workflows-automating probabilistic seismic hazard analysis calculations. In: Taylor, I.J., Deelman, E., Gannon, D.B., Shields, M. (eds.) Workflows for e-Science: Scientific Workflows for Grids, pp. 143–163. Springer, London (2007)Google Scholar
  19. 19.
    Bharathi, S., Chervenak, A., Deelman, E., Mehta, G., Mei-Hui, S., Vahi, K.: Characterization of scientific workflows. In Third Workshop on Workflows in Support of Large-Scale Science, 2008. WORKS 2008, pp. 1–10 (2008)Google Scholar
  20. 20.
    Deelman, E., Singh, G., Su, M.-H., Blythe, J., Gil, Y., Kesselman, C., Mehta, G., Vahi, K., Berriman, G.B., Good, J., Laity, A., Jacob, J.C., Katz, D.S.: Pegasus: a framework for mapping complex scientific workflows onto distributed systems. Sci. Progr. 13(3), 219–237 (2005)Google Scholar
  21. 21.
    Duncan, A.B., Patrick, R.B., Alexander, D., Junwei, C., Ben, J., John, M.: A Case Study on the Use of Workflow Technologies for Scientific Analysis: Gravitational Wave Data Analysis, pp. 39–59. Springer, London (2007)Google Scholar
  22. 22.
    Jonathan, L., Hidayat, T., Miron, L., Matthew, K.W.: High-throughput, kingdom-wide prediction and annotation of bacterial non-coding rnas. PLoS ONE 3(9), 1–12 (2008)Google Scholar
  23. 23.
    Calheiros, R.N., Ranjan, R., Beloglazov, A., De Rose, C.A.F., Buyya, R.: CloudSim: a toolkit for modeling and simulation of cloud computing environments and evaluation of resource provisioning algorithms. Software 41(1), 23–50 (2011)Google Scholar
  24. 24.
    Mann, H.B., Whitney, D.R.: On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. The Annals of Mathematical Statistics 18(1), 50–60 (1947)MathSciNetCrossRefzbMATHGoogle Scholar
  25. 25.
    Hondo, F., Wercelens, P., da Silva, W., Castro, K., Santana, I., Walter, M.E., AraÃ\(^{\circ }\)jo, A., Holanda, M., Lifschitz, S.: Data provenance management for bioinformatics workflows using nosql database systems in a cloud computing environment. In: 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 1929–1934 (Nov 2017)Google Scholar
  26. 26.
    Khawar, H., Kamran, M.: Reproducibility of scientific workflows execution using cloud-aware provenance (recap). Computing (Apr 2018)Google Scholar
  27. 27.
    Navendu, J., Ishai, M., Ohad, S.: On-demand, spot, or both: dynamic resource allocation for executing batch jobs in the cloud. Technical report (March 2014)Google Scholar
  28. 28.
    Xiao, Z., Liang, P., Tong, Z., Li, K., Khan, S.U., Li, K.: Self-adaptation and mutual adaptation for distributed scheduling in benevolent clouds. Concurr. Comput. 29(5), 1–12 (2017)CrossRefGoogle Scholar
  29. 29.
    Duggan, M., Duggan, J., Howley, E., Barrett, E.: A network aware approach for the scheduling of virtual machine migration during peak loads. Clust. Comput. 20(3), 2083–2094 (2017)CrossRefGoogle Scholar
  30. 30.
    Naghmeh, D., Saeed, S.: Learning-based dynamic scalable load-balanced firewall as a service in network function-virtualized cloud computing environments. J. Supercomput. 72(4), 1342–1362 (2018)Google Scholar

Copyright information

© Springer Science+Business Media, LLC, part of Springer Nature 2019

Authors and Affiliations

  1. 1.ITIC-CONICETUniversidad Nacional de Cuyo (UNCuyo)MendozaArgentina
  2. 2.ITICUniversidad Nacional de Cuyo (UNCuyo)MendozaArgentina

Personalised recommendations