Scheduling conditional task graphs with deep reinforcement learning

Anton Debner, Maximilian Krahn, Vesa Hirvisalo
Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL}), PMLR 233:46-52, 2024.

Abstract

Industrial applications often depend on costly computation infrastructures. Well optimised schedulers provide cost efficient utilization of these computational resources, but they can take significant effort to implement. It can also be beneficial to split the application into a hierarchy of tasks represented as a conditional task graph. In such case, the tasks in the hierarchy are conditionally executed, depending on the output of the earlier tasks. While such conditional task graphs can save computational resources, they also add complexity to scheduling. Recently, there has been research on Deep Reinforcement Learning (DRL) based schedulers, but they mostly do not address conditional task graphs. We design a DRL based scheduler for conditional task graphs in a heterogeneous execution environment. We measure how the probabilities of a conditional task graph affects the scheduler and how these adverse effects can be mitigated. We show that our solution learns to beat traditional baseline schedulers in a fraction of an hour.

Cite this Paper

BibTeX
@InProceedings{pmlr-v233-debner24a, title = {Scheduling conditional task graphs with deep reinforcement learning}, author = {Debner, Anton and Krahn, Maximilian and Hirvisalo, Vesa}, booktitle = {Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL})}, pages = {46--52}, year = {2024}, editor = {Lutchyn, Tetiana and Ramírez Rivera, Adín and Ricaud, Benjamin}, volume = {233}, series = {Proceedings of Machine Learning Research}, month = {09--11 Jan}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v233/debner24a/debner24a.pdf}, url = {https://proceedings.mlr.press/v233/debner24a.html}, abstract = {Industrial applications often depend on costly computation infrastructures. Well optimised schedulers provide cost efficient utilization of these computational resources, but they can take significant effort to implement. It can also be beneficial to split the application into a hierarchy of tasks represented as a conditional task graph. In such case, the tasks in the hierarchy are conditionally executed, depending on the output of the earlier tasks. While such conditional task graphs can save computational resources, they also add complexity to scheduling. Recently, there has been research on Deep Reinforcement Learning (DRL) based schedulers, but they mostly do not address conditional task graphs. We design a DRL based scheduler for conditional task graphs in a heterogeneous execution environment. We measure how the probabilities of a conditional task graph affects the scheduler and how these adverse effects can be mitigated. We show that our solution learns to beat traditional baseline schedulers in a fraction of an hour.} }
Endnote
%0 Conference Paper %T Scheduling conditional task graphs with deep reinforcement learning %A Anton Debner %A Maximilian Krahn %A Vesa Hirvisalo %B Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL}) %C Proceedings of Machine Learning Research %D 2024 %E Tetiana Lutchyn %E Adín Ramírez Rivera %E Benjamin Ricaud %F pmlr-v233-debner24a %I PMLR %P 46--52 %U https://proceedings.mlr.press/v233/debner24a.html %V 233 %X Industrial applications often depend on costly computation infrastructures. Well optimised schedulers provide cost efficient utilization of these computational resources, but they can take significant effort to implement. It can also be beneficial to split the application into a hierarchy of tasks represented as a conditional task graph. In such case, the tasks in the hierarchy are conditionally executed, depending on the output of the earlier tasks. While such conditional task graphs can save computational resources, they also add complexity to scheduling. Recently, there has been research on Deep Reinforcement Learning (DRL) based schedulers, but they mostly do not address conditional task graphs. We design a DRL based scheduler for conditional task graphs in a heterogeneous execution environment. We measure how the probabilities of a conditional task graph affects the scheduler and how these adverse effects can be mitigated. We show that our solution learns to beat traditional baseline schedulers in a fraction of an hour.
APA
Debner, A., Krahn, M. & Hirvisalo, V.. (2024). Scheduling conditional task graphs with deep reinforcement learning. Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL}), in Proceedings of Machine Learning Research 233:46-52 Available from https://proceedings.mlr.press/v233/debner24a.html.

Related Material