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

ndistributedjava environments thelocality ofobjects plays a crucial role in determining theperformance, scalability and stability of the overallsystem. Amanual distribution ofobjects has severaldrawbacks and requires a series of assumptionswhich may not be applicable as thesystem scales.We introduce and demonstrate how aDynamicLocalizationAlgorithm can be used to place objectsin different Java Virtual Machines based on theprocessing and communication times. Using themetrics obtained by the LocalityOptimizationAlgorithm, we have designed a placement strategyfor objects and migrated them to their optimal JavaVirtual Machine. For simplicity, we use the problemstatements of Matrix Multiplication and show how arepeated execution dynamically selects the optimalJava Virtual Machine. We have used Java Partyruntime environment to demonstrate the algorithm.