Modeling For Runtime Locality Optimizations of Distributed Java Applications Using Dynamic Localization Algorithm

In d istributed j ava environments the l ocality of o bjects plays a crucial role in determining the performance, scalability and stability of the overall system. A m anual dist ribution of o bjects has several drawbacks and requires a series of assumptions which may not be applicable as the s ystem scales. We introduce and demonstrate how a D ynamic L ocalization A lgorithm can be used to place objects in different Java Virtual Machin es based on the processing and communication times. Using the m etrics obtained by the Locality O ptimization A lgorithm, we have designed a placement strategy for objects and migrated them to their optimal Java Virtual Machine. For simplicity, we use the pro blem statements of Matrix Multiplication and show how a repeated execution dynamically selects the optimal Java Virtual Machine. We have used Java Party runtime environment to demonstrate the algorithm.