Using genetic algorithms to autonomically tune the kernel
One of the next obstacles in autonomic computing is having a system self-tune for any workload. Workloads vary greatly between applications and even during an application's life cycle. It is a daunting task for a system administrator to manually keep up with a constantly changing workload. To remedy this shortcoming, intelligence needs to be put into a system to autonomically handle this process. One method is to take an algorithm commonly used in artificial intelligence and apply it to the Linux kernel.
This paper covers the use of genetic-algorithms to autonomically tune the kernel through the development of the genetic-library. It will discuss the overall designed of the genetic-library along with the hooked schedulers, current status, and future work. Finally, early performance numbers are covered to give an idea as towards the viability of the concept.