Klaudia Dussa-Zieger

Performance comparison of routing protocols using MaRS: distance-vector versus link-state

There are two approaches to adaptive routing protocols for wide-area store-and-forward networks: distance-vector and link-state. Distance-vector algorithms use <italic>O(N x e)</italic> storage at each node, whereas link-state algorithms use <italic>O(N<supscrpt>2</supscrpt>)</italic>, where <italic>N</italic> is the number of nodes in the network and <italic>e</italic> is the average degree of a node. The ARPANET started with a distance-vector algorithm (Distributed Bellman-Ford), but because of long-lived loops, changed to a link-state algorithm (SPF). We show, using a recently developed network simulator, MaRS, that a newly proposed distance-vector algorithm (ExBF) performs as well as SPF. This suggests that distance-vector algorithms are appropriate for very large wide-area networks.

Performance comparison of routing protocols under dynamic and static file transfer connections

We compare the performance of two recently proposed distance-vector algorithms (Merlin-Segall and Extended Bellman-Ford) with a link-state algorithm (SPF), under varying file transfer workload. (Unlike the traditional distance-vector algorithms, these new distance-vector algorithms do not suffer from long-lived loops.) Our comparison is done using a recently developed network simulator. MaRS. We consider both dynamic and static file transfer connections, and both uniform and hotspot distributions of source-sink pairs. Our conclusion is that Extended Bellman-Ford performs as well as SPF in terms of delay and throughput. This suggests that distance-vector algorithms are appropriate for very large wide-area networks, since their space requirements are less than that of link-state algorithms.

Improving massively data parallel system performance with heterogeneity

The authors introduce a new type of combined SIMD/MIMD (single-instruction multiple-data/multiple-instruction multiple-data) architecture called a hybrid system. The hybrid system consists of two components. The first component is massively parallel and consists of a large number of slow processors that are organized in an SIMD architecture. The second component consists of only a few fast processors (possibly only one) which are organized in an MIMD architecture. The authors contend that a hybrid system provides a means to adequately adjust to the characteristics of a parallel program, i.e., changing parallelism. They describe the machine and application model, and discuss the performance impact of such a system. Viewing the CM-2 with its front-end as a special case of a hybrid system, they substantiate the arguments and report measurements for a Gaussian elimination algorithm.<<ETX>>

Die Verteilung paralleler Programme auf Transputer

Parallele Programme werden durch ihr Speedup-Verhalten charakterisiert. Je mehr Prozessoren einem Programm zugeordnet werden, desto ( potentiell) schneller wird das Programm ausgefuhrt. Oft tritt jedoch der Effekt auf, das ab einer bestimmten Anzahl die Prozessoren nicht mehr effizient genutzt werden konnen. In einem solchen Fall konnten die uber diese Anzahl hinaus zugeteilten Prozessoren besser von einem anderen Programm genutzt werden. Berucksichtigt man diesen Sachverhalt, so entsteht bei einer gegebenen Anzahl von Prozessoren und parallelen Programmen das Problem der optimalen Prozessoraufteilung.