Below is a collection of feedback¬† received from experts who’ve reviewed technical papers and research proposals on the packet-scale paradigm:

Novelty of the Paradigm:

The proposed congestion control paradigm is a leap in the design from existing schemes.

With respect to the richness of information that hosts can get from the network as it stands, RAPID has considerably raised the bar — far beyond what I had previously thought was possible.

This chirping idea is a great conceptual breakthrough.

Use of bandwidth estimation (through Pathchirp) in Congestion-control seems a novel and worth-while approach. I would like to see more of these types of protocols.

The proposed paradigm represents a significant departure from the traditional RTT-scale protocol operation and can help scale transport protocols to ultra high-speed networks and fairly co-exist with traditional approaches.

Technical Strength of the Research:

This is a very solid project with the potential to realize meaningful advances in a topic that has not seen much progress for a relatively long time. It has a good mix of analysis and implementation.

Research in congestion control is a battered field, but this project succeeds (against all odds) in making a convincing case that it has a reasonable chance of helping make significant progress on the topic.

This research will advance knowledge and understanding within the field of congestion control in networking as well as in queuing theory, closed-loop fluid analysis, and OS design — the first open-loop and closed-loop analysis that does not ignore fine-scale behaviors.

This is a project that seeks to re-examine fundamental concepts behind transport protocols with the goal of designing and implementing solutions that can realize much higher transfer rates than currently feasible, while achieving fairness and coexistence with existing protocols, i.e., the many deployed variants of TCP.

Potential Impact of This Research:

The possible payoffs of research coming out of this project (are) significant.

Delivering on the ability to go beyond current limitations in transport protocols is certainly of importance to many high-end applications, e.g., scientific and/or medical.

Several scientific communities necessitate transfer rates in the order of terabit-per-second, while new multimedia Internet applications are expected to rise to gigabits-per-second rates. The proposed packet-scale congestion control protocol can cope with such speed with very little overhead, without affecting the performance of co-existing low-speed TCP transfers, and is RTT-fair.

This research has the ability to go beyond current limitations in the Internet, a result that would be of clear importance to the society, and particularly to the scientific and/or medical communities.

Risks (and Dealing With Them):

This is a high risk project. One can think of many-many networking phenomena that could seemingly randomly change inter packet spacing (which is the main way congestion is determined by the paradigm).

Translating the idea of a packet-scale paradigm for congestion control into a real and meaningful solution is something that has not been realized to date.

This is a high-risk research, there are too many pitfalls and too many ways that no (or rather negative) results will be produced. The potential pay-off is significant though, significant enough to warrant taking the risk.

In spite of the many risks involved, which the project is aware of, the research plan stands a meaningful chance of delivering on its goals. There are risks involved, but those are the risks inherent in any research effort, and the project does a very commendable job of addressing them.

This project is very well conceived and organized. The research agenda is quite ambitious, but substantiated with concrete initial directions.

While the three main issues raised fundamentally impact the performance gains of the proposed paradigm, which represents a risk, the proposal includes concrete plans to design statistical techniques that can help overcome the issues.

The research activities are high-risk (a lot of things could go wrong), but at the same time high-payoff. Given the careful description of most of the issues that could prevent the deployment of the paradigm, and the initial directions pointed out — the proposed research agenda is very well defined, is aware and has thought through the potential risk. Although the research agenda is quite ambitious, it has been substantiated with concrete initial directions.



Leave a Reply