Dynamic Optimizations in Linda Systems
Stefanos Kaxiras, Ioannis Schoinas · 1998
We examine schemes for optimizing a Linda implementation in a message passing environment. We show that, with these schemes, it is possible to: 1. Capitalize on the locality of operations of Linda programs by using a distributed tuple space. 2. Increase such locality with replication of tuples. 3. Take advantage of the persistent communication paths, that are present in Linda programs, by dynamically routing templates and tuples. We present the level of our success and the constrains involved in optimizing Linda in these three directions.