Multi-Objective Reinforcement Learning for Reconfiguring Data Stream Analytics on Edge Computing

Alexandre da Silva Veith; Felipe Rodrigo de Souza; Marcos Dias de Assuncao; Laurent Lefèvre; Julio C. S. dos Anjos

Multi-Objective Reinforcement Learning for Reconfiguring Data Stream Analytics on Edge Computing

Alexandre da Silva Veith ^{1, 2} Felipe Rodrigo de Souza ^{2, 1, 3} Marcos Dias de Assuncao ^{1, 2} Laurent Lefèvre ^{2, 1} Julio C. S. dos Anjos ⁴

1 AVALON - Algorithms and Software Architectures for Distributed and HPC Platforms

Inria Grenoble - Rhône-Alpes, LIP - Laboratoire de l'Informatique du Parallélisme

2 LIP - Laboratoire de l'Informatique du Parallélisme

3 ENS Lyon - École normale supérieure - Lyon

4 UFRGS - Instituto de Informática da UFRGS

Abstract : There is increasing demand for handling massive amounts of data in a timely manner via Distributed Stream Processing (DSP). A DSP application is often structured as a directed graph whose vertices are operators that perform transformations over the incoming data and edges representing the data streams between operators. DSP applications are traditionally deployed on the Cloud in order to explore the virtually unlimited number of resources. Edge computing has emerged as a suitable paradigm for executing parts of DSP applications by offloading certain operators from the Cloud and placing them close to where the data is generated, hence minimising the overall time required to process data events (i.e., the end-to-end latency). The operator reconfiguration consists of changing the initial placement by reassigning operators to different devices given target performance metrics. In this work, we model the operator reconfiguration as a Reinforcement Learning (RL) problem and define a multi-objective reward considering metrics regarding operator reconfiguration, and infrastructure and application improvement. Experimental results show that reconfiguration algorithms that minimise only end-to-end processing latency can have a substantial impact on WAN traffic and communication cost. The results also demonstrate that when reconfiguring operators, RL algorithms improve by over 50% the performance of the initial placement provided by state-of-the-art approaches.

Keywords : Data analytics Reinforce- ment learning Multi-objective Edge computing Markov decision process Monte carlo tree searc

Document type :

Conference papers

Domain :

Computer Science [cs] / Distributed, Parallel, and Cluster Computing [cs.DC]

Computer Science [cs] / Performance [cs.PF]

Complete list of metadatas

Cited literature [31 references] Display Hide Download

https://hal.inria.fr/hal-02140844
Contributor : Alexandre da Silva Veith <>
Submitted on : Wednesday, July 24, 2019 - 9:20:37 AM
Last modification on : Wednesday, July 24, 2019 - 11:20:21 AM

icpp_2019.pdf

Files produced by the author(s)

Multi-Objective Reinforcement Learning for Reconfiguring Data Stream Analytics on Edge Computing

File

Identifiers

Collections

Citation

Export

Metrics

47

2

Contact

Multi-Objective Reinforcement Learning for Reconfiguring Data Stream Analytics on Edge Computing

File

Identifiers

Collections

Citation

Export

Share

Metrics

47

2

Contact