COMPUTER SCIENCE COLLOQUIUM
Scheduling of Data Transmission in Grid Computing
Mette Gamst
DTU Management, Department of Operations Research
Technical University of Denmark
Wednesday, 17 February, 2010 at 14:00
Auditorium U47
ABSTRACT
In this talk I present some work from my PhD study at DTU Management,
Department of Operations Research at the Technical University of
Denmark.
This talk concerns scheduling of data transmission in grid computing. In
grid computing, a number of resources are available for computing large
jobs. Before a job can be executed, all job data must be present on the
resource. The considered problem is thus to schedule jobs on grid
resources such that job data arrives in time, i.e., an integrated job
scheduling and network routing problem. I consider the offline version
of the problem, which is used for emptying job queues, for advance
reservation of resources and for analyses of changes to the grid
computing system.
The offline grid scheduling problem is NP-hard. In this talk I present
greedy heuristics, meta-heuristics and an exact branch-and-price method
for the problem. While the heuristics do not always give particularly
good solution values, the exact solution method is capable of solving
all considered test instances to optimality.
Host: Jørgen Bang-Jensen
SDU HOME |
IMADA HOME |
Previous Page
Daniel Merkle