Yesterday's keynote from Anja Feldmann (Deutsche Telekom Laboratories) was about the relationship between ISPs and P2P, i.e., challenges and opportunities. In particular, there needs to be information exchanged between ISPs and P2P system which is a huge issue. In fact, the ISP does not want to provide information about their topology which would help to increase the performance of the P2P system. However, she proposed a compromise in a way that the P2P system shall provide a list of destination IPs (i.e., other peers to which a peer may connect) and the ISP returns a ranked list of destination IPs based on their topology. In this way locality is provided during peer selection without exposing the ISP's infrastructure details. Some people say that this might face scalability/complexity issues as such a query might easily include thousands of IPs. A paper about this can be found here.
From the industry session I'd like to pick up two talks and briefly summarize them. The first from Martin Stiemerling (NEC), "A Peer-to-Peer SIP System based on Service-Aware Transport Overlays", with the goal to provide a generic P2P and transport overlay system that is not bound to a certain usage (e.g., files, voice, video). In that sense the developed a Service-aware Adaptive Transport Overlay (SATO) which includes dynamic adaptation to changes (network, context). However, they only focus on voice and for IPTV it is a complexity issue of transcoding thousands of streams in parallel (no SVC used). This work has been done in the context of the Ambient Networks project. The second talk I'd like to mention was from Victor Souza (Ericsson) about "P2P traffic in ISP Network" that had a strong relationship to the keynote (see above). For live streaming they propose a time shift mechanism comprising content ingestion, content migration, content lookup, and content download. They have a prototype out there that allows for both VoD and live streaming. Furthermore, they propose to place media caches near the customer. However, there were strong doubts that this will be possible. Anyway, it is interestin to see that others have similar problems as we have and how they're trying to solve them.
Traditionally, P2P is about file sharing and has been extended to games, voice, audio, video (VoD, Live Streaming, IPTV) applications' space. At P2P'08 I've discovered some further application areas:
- P2P-based file system: IGOR File System, "The IGOR File System Demonstration" (Bernhard Amann, Benedikt Elser, Yaser Houri, Thomas Fuhrmann)
- P2P-based Wiki engine: Piki - A Peer-to-Peer based Wiki Engine (Patrick Mukherjee, Christof Leng, Andy Schürr)
- P2P-based power control system: "Towards P2P Technologies for the Control of Electrical Power Systems" (Kolja Eger, Christoph Gerdes, Sebnem Öztunali)
Next year, the conference will take place in Seattle.