Paper review : A Comparison of Mechanisms for Improving TCP Performance over
Wireless Links (BPSK97)
Reviewer : Hai Fang (hfang@acm.org)
- Goal
To analyze the individual contribution of the different machanisms for improving
TCP performance over wireless links, hence find a combination of schemes with the
best performance.
- Contribution
This paper examines and compares the effectiveness of the different schemes (belong
to the category of end-to-end, split-connection, or link-layer) and their variants
by implementing and testing the various protocols, and experimentally analyzes the
individual mechanisms and the degree of performance improved due to each.
- Main ideas
- A reliable link-layer protocol with some knowledge of TCP results in the best
performance among the different schemes.
- The performance can be imporved by shielding the TCP sender from duplicate
acknowledgement, whereas splitting the end-to-end connection at the base station
is not needed.
- SACK and ELN result in significant performance improvements.
- Evaluation
- Significance rating: 2
This paper gives a comprehensive testing on the existing TCP implementations on
wireless links, and draws some useful analysis and conclusions. (3)
The work is based on an experimental analysis instead of some more creative idea.
(-1)
- Convincing rating
The authors describe the experimental methodology and the testing environment
they used, and provide the experimental results and the analysis.
I doubt some of their assumptions, e.g., Could the simple two-hop topology
represent the realistic networks? Could the artificially generated transmission
loss reperensent the realistic data-missing on the lossy link?
- Limitation
The author didn't test the situation where the congestion and the wireless errors
coexist.
- Conclusion
The comparison of the exsiting protocols (algorithms, methodologies) may provide an
evaluation to the individual contribution of each techniques, since normally many
tricks are used together with each other in a single protocol.
10/3/01