Home > What i've been reading > Graham’s Thesis

Graham’s Thesis

October 27th, 2010 Leave a comment Go to comments

I was lucky enough to get to read Graham Williamson’s thesis draft, and it reads very well, and justifies my line of research, i.e. enhancing existing routing protocols with location information.

Graham’s Persistance Based Routing (PBR) is efficient, at the same time as being reasonably reliable. I believe it is possible to further enhance his routing protocol with location knowledge. By using his hybrid scheme (Binary Spray and Focus + PBR)  as a basis, I believe that it can be made more efficient, and increase the reliability.

pbr

His hypothesis is:

Link persistence is an effective measurement as the basis for building a practical and flexible routing protocol in human contact networks in order to maximise delivery ratio.

He lists his main contributions as follows:

  • An analysis of the metrics which form the basis for routing in delay- tolerant networks (Chapter 4).
  • A procedure for generating synthetic contact traces which include periodicity and well-defined community structure (Chapter 5).
  • A study of structural metrics for routing and the effect of community- awareness on routing performance (Chapter 6: Section 6.2).
  • A self-managing mechanism for adapting window sizes inferred from periodicities observed in the contact pattern of nodes (Chapter 6: Section 6.3).
  • A metric, and protocol based on this, for routing in human contact net- works which is based on the persistence of links (Chapter 7).
  • A hybrid degree-persistence protocol able to maintain performance with limited routing information (Chapter 7: Section 7.4).

He identifies identifies four areas for future research

  1. Augmenting his PBR scheme with a message replication phase and identifying situations that respond well to this
  2. Using composite metrics to balance competing routing demands
  3. Investigating the combination of link persistence with link prediction
  4. Considering other dissemination goals such as publish-subscribe and anycast

ContactSim

ContactSim is Graham’s powerful discrete contact event simulator, which he used to process data, run his experiments and generate results. I had a preview of this a few months ago, but have not looked at it since. However, as he has now documented some of the main aspects of it, it may be possible to adopt this simulator for our own research, by augmenting it to include a notion of location. I will contact Graham for advice on how he thinks this could be achieved. The source code may be available in the group repository, so I will start investigating this.

  1. No comments yet.
  1. No trackbacks yet.