A parallel implementation of singular value decomposition for video-on-demand services design using principal component analysis

We have developed a mathematical model for video on demand server design based on principal component analysis. Singular value decomposition on the video correlation matrix is used to perform the PCA. The challenge is to counter the computational complexity, which grows proportionally to n3, where n is the number of video streams. We present a solution from high performance computing, which splits the problem up and computes it in parallel on a distributed memory system. © The Authors. Published by Elsevier B.V.