Scheduling data redistribution in distributed data bases
Efficiently achieved redistribution of data in a distributed data base system is crucial due to the constraint that transactions may have to be suspended during the redistribution time period. This problem is also related to the routing of packets among the processors connected by an interconnection network. It is assumed that files are unit blocks of data to be transferred, each communication link in the network can transfer only one file at the time, and each node can simultaneously send and receive as many files as the number of outgoing and of in-coming links. Each file takes unit time to move along communication links. The problem is one of determining routes and a schedule for use of links by the files so as to complete the transfer in minimum time. Important cases of the problem are shown to be NP-hard; they include the following network topologies: the general graph, the tree, and the fully connected network.
- Research Organization:
- Florida Univ., Gainesville, FL (United States)
- OSTI ID:
- 5297101
- Country of Publication:
- United States
- Language:
- English
Similar Records
Deadlock-free class routes for collective communications embedded in a multi-dimensional torus network
Multiprocessor multisystem communications network