DISCS: A Distributed Coordinate System Based onRobust Nonnegative Matrix Completion
Number of Authors: 5
2016 (English)In: IEEE/ACM Transactions on Networking, ISSN 1063-6692, E-ISSN 1558-2566Article in journal (Refereed) Epub ahead of print
Many distributed applications, such as BitTorrent, need to know the distance between each pair of network hosts in order to optimize their performance. For small-scale systems, explicit measurements can be carried out to collect the distance information. For large-scale applications, this approach does not work due to the tremendous amount of measurements that have to be completed. To tackle the scalability problem, network coordinate system (NCS) was proposed to solve the scalability problem by using partial measurements to predict the unknown distances. However, the existing NCS schemes suffer seriously from either low prediction precision or unsatisfactory convergence speed. In this paper, we present a novel distributed network coordinate system (DISCS) that utilizes a limited set of distance measurements to achieve high-precision distance prediction at a fast convergence speed. Technically, DISCS employs the innovative robust nonnegative matrix completion method to improve the prediction accuracy. Through extensive experiments based on various publicly-available data sets, we found that DISCS outperforms the state-of-the-art NCS schemes in terms of prediction precision and convergence speed, which clearly shows the high usability of DISCS in real-life Internet applications.
Place, publisher, year, edition, pages
Media and Communication Technology
Research subject Mobile and Pervasive Computing
IdentifiersURN: urn:nbn:se:ltu:diva-60526DOI: 10.1109/TNET.2016.2615889OAI: oai:DiVA.org:ltu-60526DiVA: diva2:1047580