Traditional DHT (Distributed Hash Tables) abstraction distributes data items among peer nodes on a structured overlay network. This introduces a number of issues when nodes are under different administrative authorities. In this paper, we propose DGRID, a new DHT abstraction for grid resource indexing and discovery where an administrative domain stores only its own data items. This is achieved by having each unique resource type belonging to an administrative domain to join a DHT as a node with a unique identifier. Using Chord as the underlying overlay graph, we show that DGRID lookup path length is at worst comparable with traditional DHT. However, DGRID is by design resilient to node failures without the need to replicate data items.
|Cite as: March, V., Teo, Y.M. and Wang, X. (2007). DGRID: A DHT-Based Resource Indexing and Discovery Scheme for Computational Grids. In Proc. Fifth Australasian Symposium on Grid Computing and e-Research (AusGrid 2007), Ballarat, Australia. CRPIT, 68. Coddington, P. and Wendelborn, A., Eds. ACS. 41-48. |
(local if available)