
Distributed computing involves systems that can operate across network connections, using the resources of different computers in a transparent fashion. Developing applications for such environments, and managing their execution, are problems of considerable interest in the commercial sector and present many research challenges. The University of Saskatchewan is one of six universities participating with IBM Canada in the MANDAS project (the other universities are Carleton, Michigan, Queen's, Waterloo, and Western Ontario).
In general terms, Saskatchewan is interested in issues relating to system level support for distributed applications and with issues relating to the performance of distributed systems. In addition to our work on cache management in distributed file systems, we are also exploring issues relating to understanding, characterizing, and modelling the workloads of distributed applications, with a particular focus on the Open System Foundation's Distributed Computing Environment (DCE), the midware platform adopted by the MANDAS project. Our research here is directed at achieving a better understanding of the performance of distributed applications, the performance of DCE running these applications, and the performance of systems (and networks) running DCE running these applications.
As the basis for all communication in DCE, the Remote Procedure Call (RPC) protocol has a profound impact on performance in a DCE environment, not only system performance but application performance as well. We are presently looking at issues affecting the performance of DCE RPC, and the impact of DCE RPC performance on the performance of DCE and applications running on top of DCE. We also hope to build a realistic model of RPC activity in a DCE environment using insights and data obtained from monitoring ``real'' distributed applications running on ``real'' DCE-based distributed systems. To date we have published one paper ( Measurements of DCE RPC Performance in an OS/2 Environment) presenting some early measurements of DCE performance; a second paper (Monitoring DCE Runtime Performance Using Shim Technology) describes an application to DCE of a new approach we have developed to collect real-time performance data by exploiting well-defined software interfaces.
The Saskatchewan MANDAS Group at CASCON 95
(left to right: Greg Oster, computer screen, Rick Bunt, Ying Sun)
Comp
Sci. Dept |
University |