Coordinating activities in a distributed system is an open research topic. Several models have been
proposed to achieve this purpose such as message passing, publish/subscribe, workflows or tuple
spaces. We have focused on the latter model, trying to overcome some of its disadvantages. In
particular we have applied spatial database techniques to tuple spaces in order to increase their
performance when handling a large number of tuples. Moreover, we have studied how structured
peer to peer approaches can be applied to better distribute tuples on large networks. Using some of
these result, we have developed a tuple space implementation for the Globus Toolkit that can be
used by Grid applications as a coordination service. The development of such a service has been
quite challenging due to the limitations imposed by XML serialization that have heavily influenced
its design. Nevertheless, we were able to complete its implementation and use it to implement two
different types of test applications: a completely parallelizable one and a plasma simulation that is
not completely parallelizable. Using this last application we have compared the performance of our
service against MPI. Finally, we have developed and tested a simple workflow in order to show the
versatility of our service.
Identifer | oai:union.ndltd.org:unibo.it/oai:amsdottorato.cib.unibo.it:914 |
Date | 28 April 2008 |
Creators | Capizzi, Sirio <1980> |
Contributors | Ciancarini, Paolo |
Publisher | Alma Mater Studiorum - Università di Bologna |
Source Sets | Università di Bologna |
Language | English |
Detected Language | English |
Type | Doctoral Thesis, PeerReviewed |
Format | application/pdf |
Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0023 seconds