Dynamic Scheduling for Wireless
Data Center Networks
Yong Cui, Member, IEEE, Hongyi Wang, Xiuzhen Cheng, Senior Member, IEEE,
Dan Li, Member, IEEE, and Antti Yla¨-Ja¨a¨ski
Abstract—Unbalanced traffic demands of different data center applications are an important issue in designing data center networks
(DCN). In this paper, we present our exploratory investigation on a hybrid DCN solution of utilizing wireless transmissions in DCNs. Our work aims to solve the congestion problem caused by a few hot nodes to improve the global performance. We model the wireless transmissions in DCN by considering both the wireless interference and the adaptive transmission rate. Besides, both throughput and job completion time are considered to measure the impact of wireless transmissions on the global performance. Based on the model, we formulate the problem of channel allocation as an optimization problem. We also design an approximation algorithm with an approximation bound of 1/2 and a genetic algorithm to address the scheduling problem. A series of simulations are performed to evaluate and demonstrate the effectiveness of our wireless DCN scheme.
Index Terms—Data center networks, wireless communication, dynamic scheduling, evolutionary computing, genetic algorithms
the development of cloud computing, more and more data centers are built to provide various cloud applications such as search, e-mail, and distributed file systems. As the infrastructure of data centers, data center networks (DCN) are constructed to provide a scalable structure and an adequate network capacity to bear the services.
However, current DCN, which evolves from enterprise
LAN networks, comes across more and more difficulties with the growth of cloud computing. To begin with, the rapidly increasing size of data centers brings new challenges. Nowadays, large-scale data centers usually consist of thousands of servers. For traditional Ethernet solutions, expensive high-end switches and a huge number of wires are necessary to support so many servers, which leads to a lot of troubles in wiring and maintaining.
On the other hand, data center applications with unbalanced traffic distributions suffer from inadequate network capacity. Based on the traffic statistics obtained from a real-world data center, typical data center applications such as map-reduce  usually generate a traffic demand with only a few nodes being hot (i.e., these nodes need to transmit a large volume of traffic). Fig. 1 shows an example of traffic demand matrix, where darker points stand for higher traffic
. Y. Cui, H. Wang, and D. Li are with the Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China.
E-mail: firstname.lastname@example.org, email@example.com, firstname.lastname@example.org. . X. Cheng is with the Department of Computer Science, The George
Washington University, Washington, DC 20052. E-mail: email@example.com.
. A. Yla¨-Ja¨a¨ski is with the Department of Computer Science and
Engineering, Aalto University, Aaalto, Uusimaa 00076, Finland.
Manuscript received 5 Sept. 2011; revised 27 Nov. 2012; accepted 20 Dec.
2012; published online 10 Jan. 2013.
Recommended for acceptance by D. Xuan.
For information on obtaining reprints of this article, please send e-mail to: firstname.lastname@example.org, and reference IEEECS Log Number TPDS-2011-09-0597.
Digital Object Identifier no. 10.1109/TPDS.2013.5.
1045-9219/13/$31.00 ß 2013 IEEE
demands. Although the matrix is quite sparse, those hot nodes are likely to cause loss on edge links and, therefore, put off the completion of a job. Furthermore, the nondeterministic distribution of hot nodes makes it impossible to set up additional wired links in advance to relieve the congestion. As a result, the network is prone to suffer from hot spot congestion, which cannot be