RNC: A high-precision Network Coordinate System

Abstract

Network Coordinate System (NCS) has drawn much attention over the past years thanks to the increasing number of large-scale distributed systems that require the distance prediction service for each pair of network hosts. The existing schemes suffer seriously from either low prediction precision or unsatisfactory convergence speed. In this paper, we present a novel distributed network coordinate system based on Robust Principal Component Analysis, RNC, that uses a few local distance measurements to calculate high-precision coordinates without convergence process. To guarantee the non-negativity of predicted distances, we propose Robust Nonnegative Principal Component Analysis (RUN-PACE) which only involves convex optimization, consequently resulting in low computation complexity. Our experimental results indicate that RNC outperforms the state-of-the-art NCS schemes.

NADeMaC: A simple non-negative decentralized completion algorithm for Internet latency matrix

“…Thus the accuracy of the non-negative preserving algorithms cannot be improved obviously. We tend to agree with the conclusion drawn by [16]: The meaning of the non-negative ensuring scheme is limited to the 38 reasonable physical meaning, and the applications In some specific algorithms. It can be seen that the traditional non-negative ensuring scheme drops the estimation accuracy slightly.…”