Home > Published Issues > 2017 > Volume 12, No. 7, July 2017 >

Fault Tolerance Improvement for Cloud Data Center

Humphrey Emesowum, Athanasios Paraskelidis, and Mo Adda
School of Computing, University of Portsmouth PO1 3HE, United Kingdom

Abstract—One of the main research concerns of Data Center Network designs is the achievement of reasonable throughput during multiple failures. To design a fault tolerant data center network that will give operators the operational efficacy they needed to achieve such reasonable throughput is a hard nut to crack. In this paper, we proposed an improved version of fat-tree interconnection to address the issues of fault tolerance. Fat-Tree is a well-known architecture widely used in data center networks due to its congestion control and fault tolerance capabilities attributable to its availability of alternative paths from source to destination. Our focus is on cloud data center for client to server communications, thus Email and FTP applications were run separately on all the designs. Using the same number of servers and switches, we proved that our proposed Hybrid and Reversed Hybrid designs outperformed Fat-tree topology designs at multiple failures. The results show an improvement on fault tolerance capability, and cost-effectiveness for cloud data center.
Index Terms—Fat-Tree, fault tolerance, cloud data center,graceful performance degradation, reversed hybrid

Cite: Humphrey Emesowum, Athanasios Paraskelidis, and Mo Adda, "Fault Tolerance Improvement for Cloud Data Center," Journal of Communications, vol. 12, no. 7, pp. 412-418, 2017. Doi: 10.12720/jcm.12.7.412-418.