Static worksharing strategies for heterogeneous computers with unrecoverable failures

Authors:
Anne Benoit;Yves Robert;Arnold Rosenberg;Frédéric Vivien
Affiliations:
Ecole Normale Supérieure de Lyon, France and LIP, UMR, ENS, CNRS, INRIA, UCBL, Lyon, France;Ecole Normale Supérieure de Lyon, France and LIP, UMR, ENS, CNRS, INRIA, UCBL, Lyon, France;Colorado State University, Fort Collins;INRIA, France and LIP, UMR, ENS, CNRS, INRIA, UCBL, Lyon, France
Venue:
Euro-Par'09 Proceedings of the 2009 international conference on Parallel processing
Year:
2009

Citing 11
Cited 0

Making commitments in the face of uncertainty: how to pick a winner almost every time (extended abstract)

STOC '96 Proceedings of the twenty-eighth annual ACM symposium on Theory of computing
Scheduling with unexpected machine breakdowns

Discrete Applied Mathematics
Optimal Schedules for Cycle-Stealing in a Network of Workstations with a Bag-of-Tasks Workload

IEEE Transactions on Parallel and Distributed Systems
MPI: The Complete Reference

MPI: The Complete Reference
Scheduling Divisible Loads in Parallel and Distributed Systems

Scheduling Divisible Loads in Parallel and Distributed Systems
On Optimal Strategies for Cycle-Stealing in Networks of Workstations

IEEE Transactions on Computers
Ten Reasons to Use Divisible Load Theory

Computer
Efficient collective communication in distributed heterogeneous systems

Journal of Parallel and Distributed Computing
Divisible Load Theory: A New Paradigm for Load Scheduling in Distributed Systems

Cluster Computing
Scheduling Divisible Loads on Star and Tree Networks: Results and Open Problems

IEEE Transactions on Parallel and Distributed Systems
Static strategies forworksharing with unrecoverable interruptions

IPDPS '09 Proceedings of the 2009 IEEE International Symposium on Parallel&Distributed Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

One has a large workload that is "divisible" (its constituent work's granularity can be adjusted arbitrarily) and one has access to p remote computers that can assist in computing the workload. How can one best utilize the computers toward this end? Two features complicate this question. First, the remote computers may differ from one another in speed. Second, each remote computer is subject to interruptions of known likelihood that kill all work in progress on it. One wishes to orchestrate sharing the workload with the remote computers in a way that maximizes the expected amount of work completed, given the risk of interruptions. We consider three versions of the preceding problem. Two versions envision heterogeneous computing resources: the remote computers may differ from one another in speed; one version envisions homogeneous computing resources: the remote computers are identical. One of the heterogeneous versions ignores communication costs (i.e., assumes that they are negligible); the other two versions account explicitly for communication costs. We provide exact expressions for the optimal work expectation for all three versions of the problem. For the most general version (heterogeneous resources, with communication costs), we provide a recurrence for computing this expectation; for the other two versions, we provide closed-form expressions.