The Harness metacomputing framework is a reliable and flexible environment for distributed computing. A shortcoming of the system is that services are dependent on a name service (a single point of failure) where all Harness Distributed Virtual Machines are registered. Thus, there is a need to design and implement a more reliable name service. This paper describes the Harness Distributed Name Service (HDNS) which aims to address this shortcoming. Section 2 outlines the role of the name service in Harness. Section 3 extends this discussion by describing the design of the HDNS. Finally, in sections 4 and 5, we present the services fault-tolerance mechanisms and give our conclusions.
展开▼
