Summary: Using a Genetic Algorithm to Tune openMosix Load Balancing
Alan Ritter, Michael Meehan
Computer Science, Western Washington University
Our goal is to make improvements to the openMosix load balancing algorithm by param-
eterizing various constants, then searching this space for a configuration which maximizes
performance. A Genetic Algorithm (GA) has been implemented using a fitness function
which measures throughput. As this work is ongoing, we intend to add more parameters to
the search space, and incorporate more information (e.g. network traffic and response time)
into the performance measure.
· openMosix is an open source version of the MOSIX distributed perating system.
· It is distributed as patch to the Linux kernel, and a set of user-level utilites.
· Provides transparent process migration. Processes are automatically migrated to a node
where resources (such as CPU time and memory) are available.
· A process' system calls are routed back to it's home node over the network as necissary
via Remote Procedure Calls.
The openMosix Load Balancing Algorithm
The load balancing algorithm in MOSIX is: