Home > Ideas, projects, what i've been doing > Understanding ContactSim

Understanding ContactSim

November 3rd, 2010 Leave a comment Go to comments

In the last post, I mentioned Graham’s simulator, ContactSim, which he descibes as follows:

ContactSim is a discrete-time event simulator which is specifically designed to facilitate the playback of contact traces. It is flexible, allows simple addition of new datasets and swapping of existing datasets for the underlying contact patterns, and has a powerful XML configuration language.

I have managed to have a look through it, and with the help of Graham’s appendices notes and emails, I have managed to get it working, and have started to understand how it works. Graham mentioned that he has some ideas about how I might be able to adapt it to also consider location, and automatically run the kind of simulations we have been thinking about.

Grahams Appedix A lists in more detail how the simulator works. He suggests adding a LocationUpdate event, which informs nodes (aka processes in ContactSim) of any changes in their location, which can be used by the routing algorithms.

As a test, I formatted the co-location data that we generated from the SocialSensing study into the CSV format used in the simulator (I need to get a bit more information about the correct way to do this for Graham, but for the time being I copied the formatting of the MIT Reality Mining data format used by Graham), and ran two algorithms over it, Persistance Based Routing  and Prophet.  Whilst I am still trying to work out what the results mean(!) I produced the two plots below:

delivery_ratio

This plot simply shows the delivery ratio of the PBR prototcol using the MIT dataset with 103 users, vs the Social Sensing Study data with 26 users. It covers different time periods (both of 1  month), but clips some data at the end of the MIT dataset. However, it does show that the datasets are in some way comparable.

I am continuing to investigate the software, and will speak with Graham (who is in Canada until Christmas) about how best to adapt the software for our needs. It would seem wasteful not to adopt this software which does such a similar thing, without first testing its feasability. Another simulator mentioned by Graham is the ONE simulator, which I have not yet looked at.

What I am also planning to do, is to find out whether the MIT Dataset contains some notion of location, even if that is just cell tower data, that way, it would make it easier to compare any future location based DTN algorithm that we come up with, to Graham’s work, and any other work that uses this dataset.

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