An Architecture for a Multi-threaded Harness Kernel

Abstract. Harness is a reconÞgurable, heterogeneous distributed metacomputing framework for the dynamic conÞguration of distributed virtual machines, through the use of parallel ”plug-in” software components. A parallel plug-in is a software module that exists as a synchronized collection of traditional plug-ins distributed across a parallel set of resources. As a follow-on to PVM, the Harness kernel provides a base set of services that plug-ins can use to dynamically deÞne the behavior of the encompassing virtual machine. In this paper, we describe the design and implementation details of an efÞcient, multi-threaded Harness core framework, written in C. We discuss the rationale and details of the base kernel components – for communication, message handling, distributed control, groups, data tables, and plug-in maintenance and function execution – and how they can be used in the construction of highly dynamic distributed virtual machines.

• Type: Conference
• Year: 2001
• Where: PVM
• Authors: Wael R. Elwasif, David E. Bernholdt, James Arthur Kohl, G. A. Geist II
• Keywords: Distributed And Parallel Computing, Harness Core Framework, Heterogeneous Distributed Metacomputing, PVM 2001, Virtual Machines