Mobility TIE Mobility TIE MobileAgent Scalability MobileAgent versus
Mobility TIE Mobility TIE Mobile-Agent Scalability Mobile-Agent versus Client-Server Performance Scalability in an Information Retrieval Task Dartmouth (Contact: Ron Peterson, rapjr@cs. dartmouth. edu) David Kotz, Robert Gray, Ronald Peterson Peter Gerkin, Martin Hofmann Niranjan Suri, Greg Hall, Paul Groth, Maggie Breedy Dartmouth College Lockheed Martin University of West Florida Military Application Scenario University of West Florida Testbed Mobile information server Server Agent Clients in the field Scenario Clients Parameters The following parameters were varied: Comparison of mobile-agent versus client-server implementations of an information retrieval task. * Number of clients: 1 client to 10 clients Goal is to reduce bandwidth utilization to be able to service more clients. * Network bandwidth shared by all clients to the server: 1, 100 Mbps Application is useful in low bandwidth environments such as wireless data links where reconfigurability via code distribution is important. * Percentage of documents found relevant: 5%, 20% Fixed document size: 4096 bytes Fixed query rate: 1 query per 2 seconds per client +/-0. 25 queries/second Results Legend BANDWIDTH 10 Mbps 1 Mbps Client-Server/Mobile-Agent Performance Ratio Latency 100 Mbps Client-Server D’Agents EMAA NOMADS Lessons Learned: Future Work: • Greater number of clients • Mobile Agent performance scales well • More bandwidths • Server design important • More agent systems • Mobile-Agents win at low bandwidth or with more clients Mobility TIE Mobility TIE
- Slides: 1