Cache-Enabled Super-Peer Overlays for Multiple Job Submission on Grids

Many types of distributed scientific and commercial applications require the submission of a large number of independent jobs. One highly successful and low cost mechanism for acquiring the necessary compute power is the "publicresource computing" paradigm, which exploits the computational power of private computers. Recently decentralized peer-to-peer and super-peer technologies have been proposed for adaptation in these systems. A super-peer protocol, proposed earlier by this group, is used for the execution of jobs based upon the volunteer requests of workers, and a super-peer overlay is used to perform two kinds of matching operations: the assignment of jobs to workers and providing workers the input data needed for job execution. This paper extends this superpeer protocol to account for a more dynamic and general scenario, in which: (i) workers can leave the network at any time; (ii) each job is executed multiple times, either to obtain better statistical accuracy or to perform parameter sweep analysis; and, (iii) input data is replicated and distributed to multiple data caches on-the-fly, in an effort to improve performance in terms of data availability, fault tolerance and execution time. A simulation study was performed to analyze the latest iteration of the super-peer protocol and specifically evaluate the new features