Intelligent Replication Management for HDFS Using Reinforcement Learning

TL;DR

This paper explores the application of reinforcement learning to manage data replication in HDFS, demonstrating comparable or better performance than traditional heuristics, despite current limitations in scalability.

Contribution

It introduces a reinforcement learning approach for HDFS replication management, highlighting its potential as an alternative to existing heuristics.

Findings

RL model performs comparably or better than heuristics

Experiments show potential despite scalability limitations

RL offers a promising direction for system management

Abstract

Storage systems for cloud computing merge a large number of commodity computers into a single large storage pool. It provides high-performance storage over an unreliable, and dynamic network at a lower cost than purchasing and maintaining large mainframe. In this paper, we examine whether it is feasible to apply Reinforcement Learning(RL) to system domain problems. Our experiments show that the RL model is comparable, even outperform other heuristics for block management problem. However, our experiments are limited in terms of scalability and fidelity. Even though our formulation is not very practical,applying Reinforcement Learning to system domain could offer good alternatives to existing heuristics.