Java is generally considered as a slow language when it comes to High Performance Computing. Standard performance tests in a High Performance Computing scenario, against languages like C and FORTRAN yield that Java does perform below par while executing parallel computational logic. This can be owed to slower compilation time of complex logic on a standard JVM, overhead caused due to multithreading references, inability in identifying live objects from the garbage collector for reuse, and unsafe object handling in Remote Method Invocations over a cluster environment. The paper provides a suggestive approach to customizing certain extensions on Java to optimize performance on a High Performance Computation System.