Font Size: a A A

Research Of Exchanged Folded Hypercube Network Structures

Posted on:2015-01-20Degree:MasterType:Thesis
Country:ChinaCandidate:Y LiFull Text:PDF
GTID:2298330467486277Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Topology structure is one of the most important design issues for networks. It has been found that simple topology like linear, ring, star, even the binomial tree is unavailable as the amount of nodes and data increase rapidly.In recent years, hypercube type networks have received much attention due to many of these attractive properties in massive parallel super computing systems.In this paper, two innovations are:l.we present and analyze a new hypercube type structure called the Exchanged Folded Hypercube (EFH), this hypercube network obtained by introducing some extra links to the exchanged hypercube,each node is connected to its farthest node via an edge(link) hereafter referred to as complementary edge. Compared to EH, EFH holds some features such as halved diameter of its parent topologies, lower cost factor and constant node degree. For the new proposed design, we develop the optimal routing algorithms which are simple and easy to implement. Furthermore,we provides a concise comparison with the exchanged hypercube and hypercube,exchanged hypercube,folded hypercube based on the some defined network parameters. Besides,EFH has prominent cost factor,shorter delay,less message traffic density.2.Load balancing is a significant problem which can not be ignored. In interconnection network or parallel systems, if a node breaks down because of over burden, the performance of the whole system will be affected correspondingly,even the whole system will breaks down. We first proposes a load balancing algorithm for the Exchanged Folded Hypercube structure based on the classical DE-based algorithm.Then, considered under a fault-tolerant environment,Proposes the load balancing algorithm of has m(D<m<2D-3) fault tolerance, where m is fault links number of hypercube,D is fault diameter of structure. Meanwhile, The algorithm time complexity is O(mn).
Keywords/Search Tags:Hypercube, Load Balance, Parallel Compute, Topology
PDF Full Text Request
Related items