Large Scale Peer to Peer Performance Evaluations, with Gauss-Jordan Method as an Example

Abstract

This paper presents a large scale block-based Gauss-Jordan algorithm to invert very large dense matrices. This version proposes to exploit peer-to-peer (P2P) platforms with increasingly large sets of distributed heterogeneous resources. We assume that we have access to a scheduler that proposes strategies allowing data nailing and data migration anticipation heuristics. Under given hypotheses, we present the up bounds of theoretical evaluation results, using different P2P platforms, with sufficient number of peers interconnected by different networks. Nevertheless, we discuss that, in these cases, the classical evaluation model is not well-adapted to this P2P computing paradigm for large scale scientific applications.