A Resilient Architecture Oriented to P2P Computing

This paper proposes a distributed computing architecture using the P2P paradigm. Our proposal gathers the peers into markets according to their computational resources. Each market is arranged in an N-tree and the trees are linked by a Bruijn graph. The tree topology allows efficient searching of available resources in a specific market, while Bruijn provides good scalability as search complexity does not depend on the number of markets. A Hilbert function is used to arrange markets in one ordered and mono-dimensional space. This way, the proposed architecture exploits the Bruijn and N-tree topologies together with the Hilbert function in a completely distributed, scalable and fault tolerant way. Likewise, the paper describes the main operations involved in the peer management (insertion, maintenance and output of peers) and analyses its reliability. The main parameters associated with our proposal are discussed and tuned by means of simulation. A look-up query mechanism of multiple attributes for exact and approximate queries is also introduced over this architecture. The performance of our proposal was analyzed by means of simulation in relation to the widely used Chord overlay and the Baton algorithm.