Cluster Interconnection Networks Research Articles

A network performance sensitivity metric for parallel applications

Excessive run time variability of parallel application codes on commodity clusters is a significant challenge. To gain insight into this problem, our earlier work developed tools to emulate parallel applications (PACE) by simulating computation and using the cluster's interconnection network for communication, and further study parallel application run time sensitivity effects to controlled network performance degradation (PARSE). This work expands our previous efforts by presenting a metric derived from PARSE test results conducted on several widely used parallel benchmarks and application code fragments. The metric suggests that a parallel application's sensitivity to network performance variation can be quantified relative to its behaviour in optimal network performance conditions. Ideas on how this metric can be useful to parallel application development, cluster system performance management and system administration are also presented.

Future generation supercomputers II

In part-I, a novel multi-core node architecture was proposed which when employed in a cluster environment would be capable of tackling computational complexity associated with wide class of applications. Furthermore, it was discussed that by appropriately scaling the architectural specifications, Teraops computing power could be achieved at the node level. In order to harness the computational power of such a node, we have developed an efficient application execution model with a competent cluster architectural backbone. In this paper we present the novel cluster paradigm, dealing with operating system design, parallel programming model and cluster interconnection network. Our approach in developing the competent cluster design revolves around an execution model to aid the execution of multiple applications simultaneously on all partitions of the cluster, leading to cost sharing across applications. This would be a major initiative towards achieving Cost-Effective Supercomputing.

Conflict-free channel set assignment for an optical cluster interconnection network based on rotator digraphs

Recently a class of scalable multi-star optical networks is proposed in [2]. In this class of networks nodes are grouped into clusters. Each cluster employs a separate pair of broadcast and select couplers. The clusters are interconnected via fiber links according to a regular topology. Self links are provided to enable connectivity among nodes in the same cluster if the cluster size is more than one. These networks can efficiently combine time and/or wavelength division with direct space division. The key design issue for these networks is the optimal conflict-free channel set assignment to the output clusters for a given cluster interconnection topology. Such conflictfree channel assignment problem has been studied for various cluster interconnection topologies [1, 2, 9–11]. In this paper, we propose the rotator digraph [5] as the cluster interconnection topology as it possesses many attractive properties. We will give an optimal conflict-free channel set assignment for this new interconnection topology.