Technical reports

Resource management in a distributed computing system

Daniel Hammond Craft

116 pages

This technical report is based on a dissertation submitted March 1985 by the author for the degree of Doctor of Philosophy to the University of Cambridge, St John’s College.

DOI: 10.48456/tr-73

Abstract

The Cambridge Distributed System, based on the Cambridge Ring local area network, includes a heterogeneous collection of machines known as the processor bank. These machines may run network servers, or may be loaded with services and allocated to users dynamically. The machines and the variety of services they can support (eg. different operating systems, compilers, formatters) are viewed as resources available to other components of the distributed system.

By using a processor bank, two fundamental limitations of the personal computer approach to distributed computing can be overcome: responsiveness for computation-intensive tasks is not limited by the single, personal machine because tasks may expand into processor bank machines as necessary; and applications are not limited to the operating system or languages available on the personal computer because all of the systems or languages which run on processor bank machines are at the users disposal, both for implementing new applications and for importing applications from other systems. Resource management is seen as one of the four areas which must be addressed to realize these advantages.

The resource management system must match client requirements for resources to those resources which are available on the network. To do this it maintains two data bases: one contains information describing existing resources, and the other contains information indicating how to obtain resources from servers or have them constructed from existing subresources by fabricators. The resource management system accepts resource requirements from clients and picks from the alternatives in these data bases the “best” match (as defined by the resource management policy).

The resource management issues addressed include resource description, location and allocation, construction, monitoring and reclamation, authentication and protection, and policy. The design and implementation of two resource management servers is discussed.

Full text

PDF (6.6 MB)

BibTeX record

@TechReport{UCAM-CL-TR-73,
  author =	 {Craft, Daniel Hammond},
  title = 	 {{Resource management in a distributed computing system}},
  url = 	 {https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-73.pdf},
  institution =  {University of Cambridge, Computer Laboratory},
  doi = 	 {10.48456/tr-73},
  number = 	 {UCAM-CL-TR-73}
}