Construct optimized overlay multicast via preferential random walk

Tree depth and load balancing are two main metrics in overlay multicast network. Optimizing the two metrics with lightweight overhead is important for live media streaming. This paper proposes one scheme to construct optimized overlay multicast with short tree depth and load balancing via short random walk. The key idea is the preferential random walk based on fitness function in which the tree depth and load balancing metrics are defined as parameters with weighted coefficients. Simulations and experiments show that the fitness function is valid and optimized overlay network could be constructed via preferential random walk. We also find the local and global optimized results occur at some middle value of coefficient between 0 and 1, which is not consist with our intuitions that optimized result with single metric should occur at the boundary of coefficient(0 or 1.0).