Download presentation
Presentation is loading. Please wait.
Published byΣάπφιρα Κανδάκη Γούσιος Modified over 6 years ago
1
Agent-based Resource Management for Grid Computing
Junwei Cao University of Warwick May 2002
2
Introduction Ten years research on high performance computing at Warwick: Performance evaluation Local grid scheduling Global grid management Collaborations with: NASA Ames Research Centre Los Alamos National Laboratory IBM T J Watson Research Centre
3
Agents in Context Grid Users PACE Application Tools
Global Grid Management - Agents Local Grid Schedulers PACE Performance Evaluation Engine Grid Resources PACE Resource Tools
4
Challenges Scalability. A grid has the potential to encompass a large number of high performance computing resources. Adaptability. A grid is a dynamic environment where performance of grid resources are changing over time.
5
A4 Methodology Agent structure Agent hierarchy Service advertisement
Communication layer Decision-making layer Local management layer Agent hierarchy Service advertisement Service discovery Agent capability tables A User A
6
Optimisation Strategies
Advertisement Data-push & data-pull Periodic & event-driven Discovery Local services Services in ACTs Lower or upper agents Optimisation Modelling Simulation M A User A A A A
7
A Case Study sweep3d fft improc closure jacobi memsort cpi S1 S2 S4 S3
(SGIOrigin2000, 16) S2 S4 (SunUltra10, 16) S3 S5 (SunUltra5, 16) S6 S12 (SunSPARCstation2, 16) S11 S8 (SunUltra1, 16) S7 S10 S9 sweep3d fft improc closure jacobi memsort cpi
8
Experiment 1 FIFO
9
Experiment 1 FIFO
10
Experiment 2 FIFO
11
Discovery Speed & Efficiency
No advertisement: Low speed Low efficiency Reasonable advertisement: High speed High efficiency Discovery efficiency (*100) Too much advertisement: Very high speed Very low efficiency
12
Conclusions A hierarchy of homogenous agents
Reconfigurable using different optimisation strategies Step-by-step service advertisement and discovery among agents Quantitative performance evaluation of agent behaviours
13
Future Work Use of heuristic iterative algorithms for local grid scheduling Integrating agents with grid tools, e.g. Globus, Condor, and NWS More than discovery, enabling automated negotiation and coordination Dynamic performance tuning of agent behaviours
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.