RMI-like communication for migratable software components in HARNESS

The foundation of the HARNESS experimental metacomputing system is the concept of dynamic reconfigurability of networked computing frameworks both in terms of the computers and networks that comprise the virtual machine, but also in the capabilities of the VM itself. These characteristics may be modified under user control by accessing a set of Java based kernels and via an object oriented "plug-in" mechanism that is the central feature of the system. While the usefulness of being capable to enroll and release computational resources at run-time has been already proved in the past by metacomputing systems such as PVM, the plug-in paradigm appeared in the past only in sequential environments and applications such as Netscape Navigator and Adobe Photoshop. On the contrary, HARNESS extends this concept to a parallel, distributed environment providing to users the capability to control, monitor and manage the pluggability of a distributed virtual machine as a whole. The parallel distributed plug-in model has several key advantages over statically configured metacomputing systems. At system level, the capability to reconfigure the set of services delivered by the virtual machine allows overcoming obsolescence related problems and the incorporation of new technologies. Recently, the scope of distributed programming has enlarged from clusters of computers connected to a LAN toward dynamically evolving sets of resources interconnected through the Internet. In this new scenario the capability of software components to migrate from one node of the virtual machine to another has become paramount. As a matter of fact, while in an environment where all the nodes belong to and are managed by a single entity, the disappearance of a node is relatively infrequent, in an unstructured environment such as the Internet the nodes may have a very short staying time. For this reason, it is necessary to build applications as sets of interacting migratable components. In a Java based environment such as it is HARNESS, one of the most convenient mechanism to allow components interact is RMI. RMI is a fully object oriented remote procedure call mechanism, however, it does not support component migration as it ties the reference of the remote object to its Internet address. This fact makes extremely difficult to use RMI in an environment where components need to often migrate from one node to another. In this paper we describe the user controlled task migration mechanism and the RMI-like method invocation mechanism we developed to support a programming model based on migratable components in the HARNESS system. Besides, our mechanisms are fully compliant with J2SE platform and required no changes in the base software distribution.