Download presentation
Presentation is loading. Please wait.
Published byJeremy Chambers Modified over 9 years ago
1
Application-level Scheduling Sathish S. Vadhiyar Credits / Sources: AppLeS web pages and papers
2
Introduction Everything about system is evaluated in terms of its impact on the application AppLeS – application-specific metacomputing scheduling agent Each application has its own AppLeS AppLeS designs and implements an adaptive application-specific schedule Application-centric scheduling customized to reflect application resource usage
3
Doctrines of AppLeS Both application-specific and system-specific information are needed for good schedules Performance depends on the application’s own performance criteria The distances between resources depend on how the application uses them Dynamic information to assess system state Predictions are accurate only within a particular lifetime A schedule is only as good as underlying prediction
4
Architecture Coordinator Resource Selector Planner Performance Estimator Actuator
6
General AppLeS Strategy
7
AppLeS with Jacobi The problem: Appropriate partitioning strategy to balance processor efficiencies and communication overheads, i.e. deriving partitions to obtain resource performance
8
Deriving Partitions for Jacobi Notations Per-processor execution time The goal
9
Deriving Partitions for Jacobi Communication time Soultion: system of linear equations by Gaussian Elimination
10
NWS in Jacobi
11
Resource Selection and Scheduling
13
AppLeS Benefits - scheduling
14
AppLeS Benefits – partitioning and memory usage
15
AppLeS Benefits – Partitioning and Memory Usage
16
References The AppLeS Project: A Status Report by Fran Berman and Rich Wolski. from Proceedings of the 8th NEC Research Symposium, Berlin, Germany, May 1997. Application-Level Scheduling on Distributed Heterogeneous Networks by Fran Berman, Richard Wolski, Silvia Figueira, Jennifer Schopf, and Gary Shao from Proceedings of Supercomputing 1996
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.