Probabilistic Backfilling

  • Avi Nissimov
  • Dror G. Feitelson
Part of the Lecture Notes in Computer Science book series (LNCS, volume 4942)


Backfilling is a scheduling optimization that requires information about job runtimes to be known. Such information can come from either of two sources: estimates provided by users when the jobs are submitted, or predictions made by the system based on historical data regarding previous executions of jobs. In both cases, each job is assigned a precise prediction of how long it will run. We suggest that instead the whole distribution of the historical data be used. As a result, the whole backfilling framework shifts from a concrete plan for the future schedule to a probabilistic plan where jobs are backfilled based on the probability that they will terminate in time.


Hide Markov Model User Estimate Idle Processor Runtime Distribution Previous Execution 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Lifka, D.: The ANL/IBM SP scheduling system. In: Feitelson, D.G., Rudolph, L. (eds.) JSSPP 1995. LNCS, vol. 949, pp. 295–303. Springer, Heidelberg (1995)Google Scholar
  2. 2.
    Mu’alem, A.W., Feitelson, D.G.: Utilization, predictability, workloads, and user runtime estimates in scheduling the IBM SP2 with backfilling. IEEE Trans. Parallel and Distributed systems 12(6), 529–543 (2001)CrossRefGoogle Scholar
  3. 3.
    Feitelson, D.G., Mu’alem Weil, A.: Utilization and predictability in scheduling the IBM SP2 with backfilling. In: International Parallel Processing Symposium, Number 12, pp. 542–546 (1998)Google Scholar
  4. 4.
    Zotkin, D., Keleher, P.J.: Job-length estimation and performance in backfilling schedulers. In: International Symposium on High Performance Distributed Computing, Number 8 (1999)Google Scholar
  5. 5.
    Chiang, S.H., Arpaci-Dusseau, A., Vernon, M.K.: The impact of more accurate requested runtimes on production job scheduling performance. In: Feitelson, D.G., Rudolph, L., Schwiegelshohn, U. (eds.) JSSPP 2002. LNCS, vol. 2537, pp. 103–127. Springer, Heidelberg (2002)CrossRefGoogle Scholar
  6. 6.
    Tsafrir, D., Feitelson, D.G.: The dynamics of backfilling: solving the mystery of why increased inaccuracy may help. In: IEEE International Symposium on Workload Characterization, pp. 131–141 (2006)Google Scholar
  7. 7.
    Tsafrir, D., Etsion, Y., Feitelson, D.G.: Backfilling using system-generated predictions rather than user runtime estimates. IEEE Trans. Parallel and Distributed systems 18(6), 789–803 (2007)CrossRefGoogle Scholar
  8. 8.
    Gibbons, R.: A historical application profiler for use by parallel schedulers. In: Feitelson, D.G., Rudolph, L. (eds.) JSSPP 1997. LNCS, vol. 1291, pp. 58–77. Springer, Heidelberg (1997)Google Scholar
  9. 9.
    Smith, W., Foster, I., Taylor, V.: Predicting application run times using historical information. In: Feitelson, D.G., Rudolph, L. (eds.) JSSPP 1998. LNCS, vol. 1459, pp. 122–142. Springer, Heidelberg (1998)CrossRefGoogle Scholar
  10. 10.
  11. 11.
    Feitelson, D.G.: Locality of sampling and diversity in parallel system workloads. In: 21st International Conference on Supercomputing, pp. 53–63 (2007)Google Scholar
  12. 12.
    Nissimov, A.: Locality and its usage in parallel job runtime distribution modeling using HMM. Master’s thesis, The Hebrew University (2006)Google Scholar
  13. 13.
    Feitelson, D.G., Rudolph, L., Schwiegelshohn, U., Sevcik, K.C., Wong, P.: Theory and practice in parallel job scheduling. In: Feitelson, D.G., Rudolph, L. (eds.) JSSPP 1997. LNCS, vol. 1291, pp. 1–34. Springer, Heidelberg (1997)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2008

Authors and Affiliations

  • Avi Nissimov
    • 1
  • Dror G. Feitelson
    • 1
  1. 1.Department of Computer ScienceThe Hebrew University of Jerusalem 

Personalised recommendations