Joint dimensioning of server and network infrastructure for resilient optical grids/clouds

摘要: We address the dimensioning of infrastructure, comprising both network and server resources, for large-scale decentralized distributed systems such as grids or clouds. design resulting grid/cloud to be resilient against link failures. To this end, we exploit relocation: Under failure conditions, a grid job cloud virtual machine may served at an alternate destination (i.e., different from one under failure-free conditions). thus consider requests have known origin, but assume degree freedom where they end up being served, which is case applications bag-of-tasks (BoT) type hosted machines in case. present generic methodology based on integer linear programming (ILP) that: 1) chooses given number sites topology install infrastructure; 2) determines amount capacity cater scenario failures links nodes. For latter, either failure-independent (FID) failure-dependent (FD) recovery. Case studies European-scale networks show that relocation allows considerable reduction total especially sparse topologies higher numbers sites. Adopting backup routing strategy does lead lower resource dimensions, only when adopt (especially high sites): Without exploiting relocation, potential savings FD versus FID are not meaningful.