1. Inter-Operating Grids through Delegated MatchMaking Alexandru Iosup, Dick Epema, Hashim Mohamed,Mathieu Jan, Ozan Sonmez 3 rd Grid Initiative Summer School, EU-NCIT/UPB, Bucharest, Romania, 12 July 2007 Parallel and Distributed Systems Group, TU Delft Todd Tannenbaum, Matt Farrellee, Miron Livny CS Department, U. Wisconsin-Madison

2. Agenda Context: our grid infrastructure Inter-operating grids Topology Mechanisms Conclusion

3. 3 Grid Infrastructure DAS-2 Homogeneous 5 sites, 5 clusters 200 nodes, Dual 1GHz Pentium III Myrinet and Surfnet PBS, then SGE DAS-3 Heterogeneous 4 sites, 5 clusters 260 nodes, 2 Single/Dual core AMD Opteron Myrinet/Gigabit Ethernet and Surfnet SGE, then ? Grid’5000 Heterogeneous 9 sites, 15 clusters >2000 nodes RENATER OAR GEANT link 10Gb, Jun’07

4. 4 Inter-Operating Grids More computing power Handling demand surges Good research environment

5. 5 Inter-Connecting Grids: Topology (1/4) Grids not sharing load User selects where to submit jobs Q: Can the user efficiently select resources? What is the administrative overhead for multiple accounts?

6. 6 Inter-Connecting Grids: Topology (2/4) Inter-connect any grid sites Towards full-mesh topology Q: Too many redundant paths? Routing loops?

7. 7 Inter-Connecting Grids: Topology (3/4) Create grid root, which routes between grids Fully hierarchical topology Q: Who manages the Root? Etc. Root

8. 8 Inter-Connecting Grids: Topology (4/4) Inter-connect existing grid roots Not fully hierarchical topology

9. 9 Inter-Connecting Grids: Mechanisms (1/3) Independent Koala schedulers, no load sharing Two choices: Let the user select the grid Assist the user’s grid selection (observational scheduling)

10. 10 Inter-Connecting Grids: Mechanisms (2/3) Policy: use remote grid only if local grid saturated Operation choices: Koala operates on top of local resource managers (OAR) Koala deploys its own environment, by-passing OAR

11. 11 Inter-Connecting Grids: Mechanisms (3/3) Policy: use remote grid only if local grid saturated Dynamic environment build-up Routing choices: To parent, to children, to siblings

12. 12 Inter-Connecting Grids: Mechanisms (3/3) The Delegated MatchMaking Mechanism Delegate resources, not jobs Dynamic environment build-up Peer-to-Peer exchange (negotiation?) With M. Livny, T. Tannenbaum, M. Farrellee (U. Wisc.-Madison) JM-1 obtains RM-1 from SM-1

13. 13 Inter-Connecting Grids: Mechanisms (3/3) Why should DMM work? Overall workload imbalance: normalized daily load (5:1) Temporary workload imbalance: hourly load (1000:1) Overall imbalance Temporary imbalance

14. 14 DMM High goodput Low wait time Finishes all jobs Even better for load imbalance between grids Inter-Connecting Grids: Mechanisms (3/3) Experimental Results: Performance

15. 15 Inter-Connecting Grids: Mechanisms (3/3) Experimental Results: Overhead DMM Overhead ~16% 93% more control messages Constant number of delegations per job until 80% load DMM Threshold to control o’head.

16. 16 Conclusion Inter-operating DAS-2, DAS-3, and Grid’5000 First steps accomplished Koala (inter-)operates DAS-2 and DAS-3 GRAM interface to OAR Also… analysis of Grid’5000 usage patterns On-going work Koala operating over OAR “DAS-2” image on Grid’5000: Globus, KOALA, OAR DMM implementation and experimental validation Future work Experiments DMM research, SLAs Virtualization, more application types

17. Grid’5000 and DAS-3 interconnection: scheduling issues Preserve each system usage Characterize jobs (especially for Grid’5000) Usage policies Allow simultaneous use of both testbeds One more level of hierarchy in latencies Co-allocation of jobs Various type of applications: PSAs, GridRPC, etc DAS-3DAS-3

18. KOALA over Grid’5000 and DAS-3 Goal: testing KOALA policies … … in a heterogeneous environment … with different workloads … with OAR reservation capabilities Grid’5000 from DAS-3 “Virtual” clusters inside KOALA Used whenever DAS-3 is overloaded How: deployment of DAS-3 environment on Grid’5000 DAS-3DAS-3

19. KOALA over Grid’5000 and DAS-3: how DAS-3DAS-3 DAS-3DAS-3 Lyon Orsay Rennes DAS-3DAS-3 file-server OAR DAS-3DAS-3 DAS-3DAS-3 DAS-3DAS-3 …

20. Using DAS-3 from Grid’5000 Authorize Grid’5000 users to submit jobs … via SGE directly, OARGrid or KOALA Usage policies? Deployment of environments on DAS-3 as in Grid’5000? When: during nights and week-end? DAS-3DAS-3

21. Current progress Collected traces of Grid’5000 [done] OAR tables of 15 clusters OARGrid tables LDAP database Analysis in progress KOALA over Grid’5000 [in progress] KOALA communicate with OAR for its information service [done] GRAM interface to OAR “DAS-2” image on Grid’5000: Globus, KOALA, OAR DAS-3DAS-3

22. Conclusion Use bandwidth and latency in job placements (lightpaths?) Deal with more application types (PSAs, …) A decentralized P2P KOALA Future work KOALA is a grid resource management system Support processor and data co-allocation Several job placement policies (WF, CF, CM, FCM)

23. Information Publications see PDS publication database at www.pds.ewi.tudelft.nl Web site KOALA:www.st.ewi.tudelft.nl/koala