HOW TO RUN :

-first run PubSub.java -then run workerMain.java -Enter IP and PORT ( ip:127.0.0.1 , port: 19999) -Run client enter (the same ip and port) ( there is a test in which 100 clients will be asking for worker )

• most of the changes that I made was related to RL in QlearningLoadBalancer class , Policy class and .... (read QlearningLoadBalancer.py and RL package in Server)

Reinforcement Learning

In each epoch performance is observed , some workers and clients are run and according to this observation weighs(W) are updated based on calculated rewardthen. Then new action allocates resourses to workers

Load balancing for CPU

It is one of the open issues that OS developers face , related to distribution of tasks between processing units in operating system

RL can play a significant role in associating these jobs , resulting in less time consumption