Peer-to-peer systems are very popular in current time. In peer-to-peer systems every peer contributes to maintain the infrastructure, ressources are shared and consumed. Millions of users user applications like Skype or BitTorrent. However, there is more below the surface of p2p applications.
For future peer-to-peer applications, like e.g. for peer-to-peer based social online networks or decentralized Service-oriented Architectures, the quality of service is crucial. A peer-to-peer infrastructure needs to be reliable against peer failures and provide a controlable quality similar to server-based solutions. In order to have a QoS-aware peer-to-peer infrastructure, novel monitoring and management solutions have been developed at KOM.
We can now observe the key performance indicators of a large-scale peer-to-peer system in the wild. Our monitoring approachs SkyEye.KOM gives us around 100 metrics regarding the CPU consumption at the peers, the hop count, response times, bandwidth utilization, memory usage, uptime behavior and many more. This information can be used to detect flaws, trends and shortcomings in the setup and configuration of the system.
As a next step, the user or provider of the peer-to-peer system may define quality goals, that are then automatically reached by the peer-to-peer system. For this, the monitored peer-to-peer system status is automatically analyzed using machine learning algorithms to detect effective re-configuration approaches, that are then implemented automatically. Thus a fully self-observing, self-configuring peer-to-peer system is created, that reaches and holds predefined quality goals.