SPAND: Shared PAssive Network performance Discovery
The Problem
Why Does It Matter?
Research Contributions
Questions to Answer
Our Approach:Shared Passive Measurements
Previous Work vs. SPAND
What I’m not doing
Outline
Related Work: Probing Algorithms
Probing Algorithms
Related Work: Server Selection Systems
Server Selection Systems
Related Work vs. SPAND
Benefits of Shared & Passive Measurements
Benefits of Sharing Measurements
Benefits of Sharing (Contd.)
Benefits of Passive Measurements
Benefits of Passive Measurements (Contd.)
Design of SPAND
Design of SPAND (Contd.)
Implementation Status
Heuristics at Packet Capture Host
Accuracy Metrics
Is Performance Stable?
Showing Accuracy: Methodology
Distribution of Reports
Servicing Perf Requests
Accuracy of Perf Responses
Remaining Sources of Error
Research Plan
Research Timeline
Conclusions
Future Work
Email: stemm@cs.berkeley.edu
Home Page: http://www.cs.berkeley.edu/~stemm