SLA Bound Parallel Job Generator

In order to evaluate different scheduling approaches for parallel jobs bound by a Service Level Agreement, an SLA job generator was developed as part of this project.

This SLA job generator can be used to generate parallel jobs, which, in addition to job duration and number of CPUs, they are associated with an expected earliest start time and latest finish time (constraints that are assumed to be expressed by a Service Level Agreement that binds the job). The generation of the jobs follows standard models developed by the parallel workloads community (in what concerns arrival time, number of CPUs, duration), which are enhanced with models that describe the tightness of the constraints in the SLA (namely earlier start time, latest finish time).

A demo of the SLA job generator is available here.
(please email rizos@cs.man.ac.uk if you'd like to use the demo)

Conditions of use: The demo is free to use. You can also use job sets generated by this demo for any non-commercial purpose but we take no responsibility whatsoever for any outcome resulting from this use. In case you produce any research publications using these job sets, we would appreciate it if you acknowledge the source by citing one (or both) of the following two papers (the first paper below presents an overview of this project, whereas the second paper below evaluates a number of different approaches to schedule SLAs making extensive use of job sets generated using this job generator).