Simulating P2P

[TankGirl]

Here are some pictures from a p2p simulator I have been working on.

A simulator is a great tool in p2p development as it allows you to experiment with different network architectures under different dynamic conditions without having to engage any real world testers into your early experiments. A p2p network consisting of a diverse set of randomly available nodes is such a complex dynamic system that intuition and educated guessing are simply not enough to get a good idea of how different technical solutions actually work and what sort of performance can be expected from them. The situation is further complicated by firewalls that introduce asymmetry to the behavior of the nodes. The effects of different line speeds, online times, firewalling etc. can all be simulated though and the impact of protocol and parameter changes and sudden network events can be watched in movie-like fashion in real or accelerated time.

So far I have focused my simulation experiments mostly on different connectivity models and the monitor interface as it is now allows a basic visualization of the resulting networks. The underlying object model is already somewhat more advanced as my intention is to simulate next also search traffic, bandwidth usage and spreading of content in the network. I'll provide you with more interesting pictures and other results as they become available.

The simulation-related work I have done so far has been very inspiring, providing me with a flow of fresh ideas about p2p networks and possible client features. Just to give you an example: to enhance the viewing of networks I built a simple 3D engine to allow 'flying' through the peer space and looking at it from different angles. When using the system for the first time I noticed immediately how wonderful the experience of flying in the peer space was. Once the peer space becomes a place to move in and around the present p2p way of searching for content and other peers with textual strings starts to feel old-fashioned and uninteresting. Why not to combine the thrill of virtual environments that we already love and enjoy in games to the p2p experience! You can easily imagine what exciting new possibilities this would open up for sharing, searching and socializing in a p2p environment...

Another interesting option is to use a p2p simulator as a 'wind tunnel' to hook a prototype p2p client into and to see how it behaves in a dynamic environment interacting with other nodes. This not only helps to develop a more ergonomic user interface but also allows to test how the client feels and behaves for different types of users - broadband vs. dial-up, heavy sharer vs. light sharer etc. All this of course requires a fairly detailed and sophisticated simulation model but the approach will be useful even if all the functionality has not yet been modeled into the simulator or the client.

- tg