|Title||Memory Compression Techniques for Network Address Management in MPI |
|Publication Type||Conference Paper |
|Year of Publication||2016 |
|Authors||Guo, Y, Archer, CJ, Blocksome, M, Parker, S, Bland, W, Raffenetti, K, Balaji, P |
|Conference Name||International Parallel and Distributed Processing Symposium (IPDPS) |
|Date Published||05/2017 |
|Abstract||MPI allows applications to treat processes as a logical collection of integer ranks for each MPI communicator, while internally translating these logical ranks into actual network addresses. In current MPI implementations the management and lookup of such network addresses use memory sizes that are proportional to the number of processes in each communicator. In this paper, we propose a new mechanism, called AVRankmap, for managing such logical addressing. AV-Rankmap takes advantage of logical patterns in rank mapping that most applications naturally tend to have, and it exploits the fact some aspects of the network address translation are naturally more performance critical than others. It uses this information to compress the network address management structures. We demonstrate that AV-Rankmap can achieve similar or better performance compared with that of other MPI implementations while using significantly less memory.