Workshop 3: Scheduling and Load Balancing

Programme Committee:

General Chair. Susan Flynn Hummel (IBM T.J.Watson Research Center), hummel@mono.poly.edu
Local Chair. Graham Riley (University of Manchester), griley@cs.man.ac.uk
Vice Chair. Rizos Sakellariou (University of Manchester), rizos@cs.man.ac.uk
Vice Chair. Wolfgang Gentzsch (GENIAS Software, Germany), gentzsch@genias.de

Description:

Mapping a parallel computation onto a parallel computer system is one of the most important questions for the design of efficient parallel algorithms. Especially for irregular data structures the problem of distributing the workload evenly onto parallel computing systems becomes very complex. This workshop will discuss the state of the art in this area. Besides the discussion of novel techniques for mapping and scheduling irregular dynamic or static computations onto a processor architecture, a number of open problems are of special interest for this workshop.

One is the development of partitioning algorithms used in numerical simulation, taking into account not only the cutsize but also the special characteristics of the application and their impact on the partitioning (e.g. shape of the partitions). Another topic of special relevance for this workshop concerns re-partitioning algorithms for irregular and adaptive finite element computations that minimize the movement of vertices in addition to balancing the load and minimizing the cut of the resulting new partition. A third topic in this uncomplete list of open problems is the development of dynamic load balancing algorithms that adapt themselves to the special characteristic of the underlying parallel computer, easing the development of portable applications.

Topics of interest include: