Home > Uncategorized > Meeting with Pádraig and Davide

Meeting with Pádraig and Davide

November 5th, 2010 Leave a comment Go to comments

Met initially over lunch then in the office. Pádraig had made the following observations:

I think the “Routing with Location” post is very useful for making concrete how location information might help routing. This is the key principle:
“the more locations it visits in common with the destination node, the more likely it will be to pass the message to the destination.”
To be devil’s advocate, the worry would be that the useful information entailed in location is already compiled into the existing contact information. If that is true then the uplift due to the inclusion of location criteria in the routing will not be significant.

Davide and I had been discussing this earlier in the day, and I explained my idea of measuring the importance of a location in terms of how ‘active’ it was; i.e. how many messages exchanges happen there. In this way, it is distinct from contact knowledge alone.

Davide suggested a similar mechanism, which was to measure the popularity of locations, based on, for example the number of different people that visit there. I misssed slightly the meaning when he said an instantaneous value. We talked about concept of maintaining a global table, which tracks the current popularity of locations, and routing to nodes that visit these locations most often.

We discussed the possibility of using the cell tower information contained within the MIT reality mining dataset, and suggested mechanisms for implying location from this. The first, assuming that the nodes are connected to the same network operator, simply being when two nodes are at the same cell tower T1 and T1, they are co-located, else they are not (e.g. they see T1 and T2 respectively)

Another situation to test for however, is when two nodes can ‘see’ each other, but they do not report the same cell tower (T1 ad T2). I suggested a hybrid location, e.g. T12 which is a location where two people have been seen to be connected, but were not connected to the same tower. Initially, we decided to analyse the data to see what it looks like. Pádraig suggested using a simple matrix – People over cell tower IDs with a count of how many times they occur.

One thing we didn’t factor for however, is whether a node can see multiple cell towers at one…. this will become clearer when I have a proper look at the dataset.

Incidentally, this got me thinking about a generic way to represent locations, as 1d, 2d, 3d etc. The MIT dataset cell tower locations, providing all nodes are using the same co-ordindate scheme (i.e. network operator), can be described as a 1d co-ordindate system. The ‘distance’ between points, whilst not meaningful for physical location, could be based on some metric of common movements between locations.

We also talked about ContactSim, and Pádriag agrees that we should probably use this if it is suitable. He asked me to ask Graham for a copy of his chapter that deals with his PBR protocol (Chapter 7) for him to read. We decided that I should try to replicate Graham’s results, to ensure I had a proper handle on what the simulator does.

Pádraid would also like me and Davide(?) to hold a seminar for few people, in which we at least show the results from Grahams work, and discuss the ideas we have about routing with location, I suggested about a year from now, Pádraig was thinking about 2 weeks from now, so in next week’s meeting we will fix the details for the following week.

For next time, I will have replicated Graham’s results, analysed the MIT dataset for applicability for our needs, will have started planning for the seminar, and will have started to adapt Graham’s simulator to use location information.

Categories: Uncategorized
  1. No comments yet.
  1. No trackbacks yet.