Sparse networks tolerating random faults
โ Scribed by T. Yamada; K. Nomura; S. Ueno
- Publisher
- Elsevier Science
- Year
- 2004
- Tongue
- English
- Weight
- 350 KB
- Volume
- 137
- Category
- Article
- ISSN
- 0166-218X
No coin nor oath required. For personal study only.
โฆ Synopsis
A network G * is called random-fault-tolerant (RFT) network for a network G if G * contains a fault-free isomorphic copy of G with high probability even if each processor fails independently with constant probability. This paper proposes a general method to construct an RFT network G * for any network G with N processors such that G * has O(N ) processors. Based on the construction, we also show that if G is a Cayley, de Bruijn, shu e-exchange, or partial k-tree network with N processors and M communication links then we can construct an RFT network for G with O(N ) processors and O(M log N ) communication links. Cayley networks contain many popular networks such as circulant, hypercube, CCC, wrapped butter y, star, and pancake networks.
๐ SIMILAR VOLUMES
Recurrent neural networks with full symmetric connectivity have been extensively studied as associative memories and pattern recognition devices. However, there is considerable evidence that sparse, asymmetrically connected, mainly excitatory networks with broadly directed inhibition are more consis
Broadcasting is the task of transmitting a message originated at one processor of a communication network to all other processors in the network. A minimal k-fault-tolerant broadcast network is a communication network on n vertices in which any processor can broadcast in spite of up to k line failur
We consider broadcasting in radio networks that are subject to permanent node failures of unknown location. Nodes are spread in a region in some regular way. We consider two cases: nodes are either situated at integer points of a line or they are situated in the plane, at grid points of a square or
In this paper, we propose a solution to both fault tolerance and hot-spot contention problems in multiprocessor systems with multistage interconnection networks. Combining networks are known to be effective in handling hot-spot traffic. However, the fault tolerance capability of unique-path combinin