Recent sojourn time results for Multilevel Processor-Sharing scheduling disciplines

Samuli Aalto

Helsinki University of Technology


Résumé:

Multilevel Processor-Sharing (MLPS) disciplines refer to a family of age-based scheduling disciplines introduced already decades ago. A time-discretized version of an MLPS discipline is applied in the scheduler of the traditional UNIX operating system. In recent years, MLPS disciplines have been used to study the way that packet level scheduling mechanisms impact the performance perceived at the flow level in the Internet. Inspired by this latter application, many new sojourn time results have been discovered for these disciplines in the context of the M/G/1 queue. In this presentation, we highlight some of these new results. In addition, we point out some intriguing open problems for further research.


Samuli Aalto
Helsinki University of Technology