Faculty Advisor
Dror Feitelson

Grant
$19,461- 1/1/01 (1 year)

Project/Description

Scheduling Interactive Processes in Linux

Scheduling in Unix systems has not changed much in 30 years, and relies on prioritization based on (low) CPU usage. This was originally sufficient to give good performance to interactive jobs, as they were typically I/O bound. However, today's interactive jobs use graphical interfaces and may use significant CPU resources, making it hard for the scheduler to identify them. We propose as alternative approach in which interactive processes are identified directly by their interactions with the X server. The scheduler will then be modified to give preferential treatment to the identified processes, guaranteeing a low scheduling latency and good responsiveness.