Gurtov and Floyd in this paper, discuss and study the complex interactions between wireless link characteristics and transport protocols. The TCP/IP stack has become the de facto technology in both wired and wireless networks and transport protocols and wireless technologies should be designed with a better understanding of the interplay between these mechanisms.
Researchers who are working on new transport protocol designs are in need of simulation models which capture the important/relevant characteristics. The authors claim that many of the current models have one or more of the following 4 type of problems: 1) being unrealistic, 2) having limited parameter space, 3) being overly realistic and 4) being under specified.
For example it was believed that abrupt increase in delay caused by link level errors will cause retransmission timeouts while later it was observed that because of relative high variations in delay, congestion windows are already inflated and rather than producing spurious timeouts they will cause slow recovery from an actually packet loss.
The authors indicate the following to be the main wireless link characteristics which need to be properly addressed in any such model:
The paper is a good overview of the complex relationship between link characteristics and transport protocols. Even though the layered structure has been relatively successful in separating the functions, these layers are not in any way independent of each other and a deep understanding of the interactions is required for any good design.
As authors have pointed out, cross layer techniques are one of the most effective methods for improving the efficiency of interplay between layers and it could be interesting to discuss the advantages/disadvantages of increase in deployments of cross-layer techniques in today's wired/wireless internet. I vote for keeping this paper in the syllabus.
Researchers who are working on new transport protocol designs are in need of simulation models which capture the important/relevant characteristics. The authors claim that many of the current models have one or more of the following 4 type of problems: 1) being unrealistic, 2) having limited parameter space, 3) being overly realistic and 4) being under specified.
For example it was believed that abrupt increase in delay caused by link level errors will cause retransmission timeouts while later it was observed that because of relative high variations in delay, congestion windows are already inflated and rather than producing spurious timeouts they will cause slow recovery from an actually packet loss.
The authors indicate the following to be the main wireless link characteristics which need to be properly addressed in any such model:
- Error losses and corruption
- Delay variation
- Packet reordering
- On-demand resource allocation
- Bandwidth variation
- Asymmetry in Bandwidth and Latency
The paper is a good overview of the complex relationship between link characteristics and transport protocols. Even though the layered structure has been relatively successful in separating the functions, these layers are not in any way independent of each other and a deep understanding of the interactions is required for any good design.
As authors have pointed out, cross layer techniques are one of the most effective methods for improving the efficiency of interplay between layers and it could be interesting to discuss the advantages/disadvantages of increase in deployments of cross-layer techniques in today's wired/wireless internet. I vote for keeping this paper in the syllabus.
No comments:
Post a Comment