Integer Linear Programming Models and Cycle Avoidance Methods for Media Streams Planning
|Original title:||Integer Linear Programming Models and Cycle Avoidance Methods for Media Streams Planning|
|Authors:||Pavel Troubil, Hana Rudová|
|Type:||Appeared in Conference without Proceedings|
|Keywords:||media streams planning; integer programming; cycle avoidance|
High-end collaborative tools simultaneously transmit multiple multimedia streams, each in an 1:n scheme. Bandwidth required by the streams is close to capacities of current links, and collaboration requires low latency. The Media Streams Planning Problem is a problem of routing streams in these tools. For each data source, a distribution tree is placed in the network. We optimize overall latency of all streams. Due to requests for quick call establishment, solver of the MSPP has to respond within seconds. Our aim is to improve its performance. We linearized the original CP formulation of the problem and proposed binary ILP solver to improve performance. We noticed that negative influence comes from cycle avoidance constraints. We proposed strategies which solve the problem without cycle avoidance methods, and eliminate only cycles appearing in the solutions. We also combined binary formulation with network flows constraints. This method avoids cycles and improves performance.