New Fault-Tolerant Datacenter Network Topologies

摘要

Data Center Network (DCN) topologies, being integral parts of cloud computing infrastructure, are vulnerable to failures due to the presence of huge number of servers, switches and links. One of the biggest challenges in the fault-tolerant DCN design is to provide a graceful degradation in performance in the event of a link or server failure. This paper proposes two new fault-tolerant DCN topologies, called DCell-Star and DCell-Ring. The proposed topologies are cost-effective, scalable, fault-tolerant, and enhance overall performance (throughput and latency) of the standard DCell topology. Performance evaluation of the proposed topologies is done through a simulation study, and the results are compared with the standard DCell topology. The comparison is done in terms of various metrics such as throughput, latency, and diameter. The simulation results show that the proposed topologies outperform the standard DCell topology due to the availability of multiple alternate shortest paths between pairs of servers, resulting in an improvement of about 5% in throughput even for a small-size network.