非階層網におけるV.O.D.サービスの蓄積割当て

概要

論文の詳細を見る
In a previous paper[2], we have already treated tree networks, and shown that a O(N logN) algorithm could be used for planning optimization. This cannot be generalized easily to the other network topologies, which we will study here. We will use the cost function defined in[2]. The only general result is that the optimal storage points are a subset of the nodes of the network[1]. In other words, it is never optimal to place the video servers on the lines, somewhere between the nodes. It is an integer programing problem but the constraint matrix is rather large. For the 36-node network we used in our computation, we get a matrix with 1296・1368 elements, corresponding respectively to the 36^2=1296 variables and 1296+36+36=1368 constraints, For such a size, it seems difficult to use classical integer programming techniques in an efficient way. References [1]G.Y.Handler and P.B.Mirchandany, Location on Networks-Theory and Algorithms, MIT Press, Cambridge, Mass(1979). [2]Y.Tanaka and O.Berlage:"Storage allocation in Video-on-Demand Service", Technical Report of IEICE, IN94-11(April 1994).
社団法人電子情報通信学会の論文
1994-09-26