A Dynamic-Resource-Allocation Based Flash Crowd Mitigation Algorithm for Video-on-Demand Network

Video-on-Demand (VoD) is a typical human-to-machine communication system, in which the application server resources are typically shared by multiple end users and are designed to satisfy normal peak traffic volume based on time-of-day effect. However, the event that catches the attention of a large number of people will trigger flash crowd, which causes unexpected and overloading surge of traffic. In this paper, a Dynamic-Resource-Allocation (DRA) based flash crowd mitigation algorithm for multicast and unicast hybrid VoD network is proposed, which can handle the flash crowd traffic smoothly though dynamically allocating multicast streaming resources to serve all the flash crowd requests instead of setting up unicast streaming for every request. Based on the study of the multicast and unicast hybrid cable network, video objects access model and flash crowd model, the simulation is done to show that the proposed DRA based flash crowd mitigation algorithm is a cost-effective and robust mechanism for flash crowd mitigation.