Cilk: an efficient multithreaded runtime system
Journal of Parallel and Distributed Computing - Special issue on multithreading for multiprocessors
A scalable, robust network for parallel computing
Proceedings of the 2001 joint ACM-ISCOPE conference on Java Grande
Framework for Peer-to-Peer Distributed Computing in a Heterogeneous, Decentralized Environment
GRID '02 Proceedings of the Third International Workshop on Grid Computing
SuperWeb: Towards a Global Web-Based Parallel Computing Infrastructure
IPPS '97 Proceedings of the 11th International Symposium on Parallel Processing
Sabotage-Tolerance Mechanisms for Volunteer Computing Systems
CCGRID '01 Proceedings of the 1st International Symposium on Cluster Computing and the Grid
Volunteer computing
On the Design of a Performance-Aware Load Balancing Mechanism for P2P Grid Systems
GPC '09 Proceedings of the 4th International Conference on Advances in Grid and Pervasive Computing
A self-adaptive load balancing strategy for p2p grids
ICA3PP'10 Proceedings of the 10th international conference on Algorithms and Architectures for Parallel Processing - Volume Part II
| Hi-index | 0.00 |
This paper presents the Jalapeno grid computing system. Jalapeno is implemented in Java and uses peer-to-peer technology provided by Project JXTA. The Jalapeno system consists of manager, worker and task submitter hosts. The task submitter submits a collection of tasks, a task bundle, to be processed by the system to a randomly chosen manager. The manager splits the bundle into a set of new, smaller bundles which are forwarded to equally many, randomly chosen, managers which repeat the process. Each manager has a small number of connected workers. During task bundle splitting the manager may, depending on its current load, reserve a number of tasks for its workers. Workers return the results to their managers which forward them to the task submitterThe system is self configuring: hosts volunteering their computing power will at first become workers only but will eventually become managers if they can not connect to another manager within a certain time.The major contributions of this project are: an implicit hierarchy of hosts which changes randomly over time and requires no effort to maintain, a framework for applications solving embarrassingly parallel type of problems which automatically partitions the problem into smaller sub-problems and ease of use through the use of Sun's Java Web Start technology.Two applications have been developed for the system to evaluate its performance: an RC5 key cracking application and a 3d ray-tracing application. The entire system is available for download at http://jalapeno.therning.org.