A fuzzy based hierarchical flash crowd controller for live video streaming in P2P network

A Peer-to-peer (P2P) network system has proved to be a popular and efficient approach for large-scale live video streaming and file sharing over the Internet. Such a system is always challenged by various issues and requires to be monitored and controlled for significant performance. Among them, one of the major performance issues in P2P networks is the flash crowd. When many users switch to a popular channel in a short span of time, the P2P system suffers from a serious join congestion problem, called the flash crowd problem. In such situations, systems can scale only up to a limit, both by available surplus bandwidth and intense competition among peers for their start-up, and hence, it not only decreases the efficiency but also creates a blockage of services for connected peers. Therefore, management of the flash crowd has been important and becoming paramount with the increased use of online services and the proliferation of various technologies like IoT, Edge & Mobile Computing, Crypto Currencies, and Torrenz-like services. A hierarchy-based admission control model with multiple threshold layers is proposed to mitigate the flash crowd in the P2P network paradigm. A fuzzy logic-based prediction mechanism is proposed to predict the control scheme applicable to the requested peers based on the parameters such as buffer size, bandwidth, etc. In case of the occurrence of the flash crowd, the model applies different control schemes depending upon which threshold layer it is predicted. These include saving the system resources, granting preferences to best-performing peers, dropping requests of worst-performing peers, a slight decrease in service quality, etc. We also characterize the credit score of peers, which denotes their performances. The model is simulated in OMNET++ along with the OverSim framework. Results from the simulation reveal a reduction in request drops, processing delays, and startup delays compared to other management schemes.