1248
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
Node Importance of Data Center Network Based on Contribution Matrix of Information Entropy Kai Peng, Rongheng Lin, Binbin Huang, Hua Zou, and Fangchun Yang State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China Email:
[email protected], {rhlin, huangbinbin, zouhua, fcyang} @bupt.edu.cn
Abstract—With the development of cloud computing, data center network (DCN) architectures as the core of the cloud platform received a surge of interesting from both the industry and academia. However, assessments of those new DCN architectures are mainly concentrated in load balancing, improvement of architectures and as well as some research of performance analysis in visualized environment. Moreover, none of them focus on the security in DCN architectures. In this paper, we propose contribution matrix method based on information entropy theory from the point of view of node importance which is the basic of the topology vulnerability. In addition, we conduct an experimental evaluation of the state-of-the-art Multi-rooted Tree and FiConn architectures, each respectively as a representative of the switch-centric network architecture and server-centric network architecture. Firstly, we use an undirected graph theory to describe the architecture. Secondly, we obtain the associated matrix of betweenness and degree, and calculate their weights by information entropy theory. And then, according to the definition 3 (see Section III-B); we obtain the contribution matrix for each one. Last but not least, we get the value of node importance for each node of normalized. Compared with other methods, our proposed method is effective and has a much higher accuracy. Furthermore, our method is generic and can be widely used for the new DCN architectures. Index Terms—DCN; Topology, Contribution Matrix, Multirooted Tree, FiConn
I. INTRODUCTION Cloud computing entrusts remote services with enterprises or individuals’ data, software and computation, where the customers enjoy the pay as use high quality applications and services from a shared integration pool of computing resources [1]. This significantly improves the quality of service and also reduces the cost greatly. Data center network (DCN) architectures as the core of the cloud platform received a surge of interesting from both the industry and the academia research. What is more, driven by the proliferation of cloud computing, web services [2] and development of IT facilities, DCN are experiencing a rapid growth in both size and complexity. Data center network is a network infrastructure of the data center, which make a large number of servers be connected by high-speed links and switches [3]. Furthermore, the topology also decides the specific forms of data center networks. Generally speaking, according to whether the servers have the capacity of forwarding packet by academia,
© 2013 ACADEMY PUBLISHER doi:10.4304/jnw.8.6.1248-1254
DCN Architectures can be divided into two categories, Switch-centric Architecture and Server-centric Architecture. However, the primary study of evaluation of these new architectures is limited to Map Reduce or scientific computing task, and currently some of them focus on how these architectures are affected in virtualized environments [4], however, none of existing research consider about the security of the data center network from the perspective of topology vulnerability. In this paper, we fill this void by conducting an experimental evaluation of Multi-rooted Tree and FiConn Architectures. We assess these architectures from the point view of node importance which is the basis of vulnerability of network topology. Node importance is investigated from the network survivability in network fault management which has been well studied in complex networks and communication networks. Existing methods can be divided into two categories, single node method (e.g. [5-6], to list a few), network associated method [7]. Taking shortest path method [5], node contraction method [6] and network associated method [7] for example. Determine which are the important nodes on the shortest path in the network was first proposed by Corley et al in [5], but their method unable to assess the importance of the nodes on whole networks. Node contraction method was introduced in [6]. Compared with the shortest path method, it does not need to remove the node; it preserves the integrity of the network topology, however it unable to determine the importance of symmetric nodes. In general, single node methods [5-6] ignore the changes of network performance after node is removed. In other word, the analysis for single node cannot well reflect the node importance in the real scenario. The method of network correlation based on the contribution matrix [7] was used in satellite mobile communications network which has well dynamic effects. But it does not take into account of the differences between the degree and betweenness in real network, thus the accuracy of it needs to be improved. For the consideration of inadequate of the previous ones, in this paper, we define and solve the problem of node importance in DCN for the research of vulnerability of topology. We propose a method of contribution matrix based on information entropy. Firstly, we use adjacency matrix of undirected graph to describe network topology of Multi-rooted Tree and FiConn Architectures. Secondly, we construct the evaluation matrix of betweenness and
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
degree as input parameters, and then calculate their weights by using information entropy theory. Thirdly, according to the definition 3 (see Section III-B), we get the contribution matrix. Last but not least, we obtain the value of the importance of each node by normalization， which clearly point out which are the important ones. Our contributions are summarized as follows, 1) First, we explore the problem of topology vulnerability in DCN, which provides a basis for the indepth research of offense and defense in cloud platform. 2) Second, this paper introduces the information entropy theory for the node degree and betweenness, the results accurately reflect the importance of the node. 3) Third, the experimental evaluation of Multi-rooted Tree and FiConn, each respectively as a representative of Switch-centric and Server-centric Architectures. The results show that our method is effective and can be widely used for new architectures. The remainder of this paper is organized as follows. In Section II, we introduce the background and related concepts of our mode. The main mode and algorithm process are described in Section III while Section IV presents the simulation results on both Multi-tree and FiConn Architectures and out discussion, followed by section V, we discuss related work on DCN and node importance. Finally, we conclude the paper in Section VI. II. BACKGROUND In this section, we briefly introduce the Switch-centric and Server-centric Architectures, which are the core of our paper. And then we mainly focus on the basis concepts of our method. A.Switch-centric Architectures In Switch-centric architectures, only the switches have the ability to forward packets while the servers do not have. These are widely used in traditional data center network, especially for the current data centers. The architectures are usually composed by routers and switches of two or three layers, such as Multi-rooted Architecture. For instance, new architectures such as Fattree [8], VL2 [9], Monsoon and Jellyfish which evolved from Tree.
1249
storing and computing. In the middle layer is aggregation layer, where places collection switches. The core layer is in the root of tree, where places the core switches. The main work of the layer of aggregation and core is to forward packets. Each switch in collection layer make dozens or hundreds of servers be connected in the down and core switches on the top. B. Server-centric Architectures Server-centric Architecture means that not only the switchers, but also the servers in the network have the ability of forwarding packet. Such topology with recursive hierarchy is based on composite figure theory. Specifically, the whole DCN consists of a number of basic units of recursively by certain way of interconnection. In general, a basic unit consists of a commercial switcher with a number of n points and n commercial servers. The main characteristics of this type can be summarized as follows. In this topology, the basic components are connected by inner servers while the switches are never connected. Typical architecture of this kind is Dcell [10], Bcube [11] and FiConn [12]. As can be seen from the Figure 2, we can see the structure of FiConn. Several isomorphic of (k-1)-th layer consist of the layer of k-th network in a certain interconnection way. In other word, layer of (k-1)-th makes all lower levels structure of k-th network be fully connected. C. Related Concepts of Our Mode We try to abstract of network topology vulnerability by to using graph theory and use undirected graph represent those architectures. Let (1) is named as adjacency matrix of G. Among many factors of node importance, in this paper, we choose degree and betweenness. • Definitation1.Degree The number of edges incident to the vertex of A is counted as degree. The degree of a vertex is denoted as (2). The number of edges connected to the node was treated as the basis of node importance. The more adjacent links the node has, the more important the node is. The degree only reflects the local impact of the node in the static network. Degree alone does not reflect the importance of nodes well. • Definitation2.Betweenness The betweenness of a node is shown by the expression (2) [7]. The betweenness of a vertex is denoted as . (2)
Figure 1. Multi-rooted Tree Architecture
As is shown in Figure 1, we can see that all the hosts are placed on the edge of the layer, forming the leaf nodes of the whole networks, which is responsible for
© 2013 ACADEMY PUBLISHER
where
,
shortest paths between node
is the total number of and node of
and
1250
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
is the number of these paths that go through . Be careful for that the calculation of betweenness of a node scales may be adjusted by dividing through by the number of pairs of nodes while not including . The division is done by for undirected graphs, in our paper, n (n-1)/2 and for directed graphs (n-1) (n-2), where n is the number of nodes. The betweenness reflects the ability of providing the shortest route of network communication tasks of nodes. It can be used to measure the ability of controlling network resources of the nodes. For the consideration that betweenness fully reflect the interactive capabilities of the nodes. It can be widely used for determine where is the heavy information load network nodes through the flow of information. Moreover, it is also suited to alleviate network congestion and cascading failure malicious attacks [14].
is the degree and is the In this paper, let m=2, between’s of each node, and thus R can be expressed as follow.
where, 2) The Evaluation of Entropy Weight Indicators [15] The entropy of i-th evaluation index is defined as (3). (3)
where, ,
Esepecailly, when
and
3) The Evaluation of Entropy Weights The entropy weights of the evaluation (4)
. is defined as
(4)
Figure 2. FiConn Architecture
III. ALGORITHM PROCESSES AND PSEUDO CODE Our method is based on the Information entropy theory and contribution matrix. The former one is for the calculation of the weights of both degree and degree (see Section II), and the other one (see Section II) is to reflect the interaction among each other. We give a brief introduce of our method and corresponding pseudo code in section C. A. Information Entropy Theory Considering the difference of importance between degree and betweenness, the weight of them should be seriously consideration. Taking into account the objectivity and low complex, we choose the information entropy theory. In addition, for the evaluation information all comes from the inside of system rather than the outside, thus the accuracy can be a good guarantee. The theory can be summarized as three steps. First is the construction of evaluation matrix, and then is the evaluation of entropy weight indicators, then the last one is the evaluation of entropy weights. 1) The Construction of Evaluation Matrix [13] For an evaluation of m, evaluation object of n, standard evaluation matrix R of multiple indicators can be definite as follows.
where,
, and
=1. From Equation 4, we
can see that the smaller the indicators entropy is, the greater entropy weight will be. B. Node Importance Contribution Matrix • Definitation3. NICM [7] Node Importance Contribution Matrix is called NICM for short. In an undirected network which has nodes of n has degree of , then it will but without ring. If node for each contribute to its own importance of adjacent node. Select node betweenness as NICM nodes in the importance of the initial value, the formation of HNICM may be denoted as follows (5).
(5)
After adding weight for both degree and betweenness (see section B), the new contribution matrix is represented as (6).
(6)
and respectively represent the weight of where degree and betweenness (see Section A).
© 2013 ACADEMY PUBLISHER
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
C.Algorithm and Pseudo Code Our Algorithm can be described as four steps. 1) Firstly, we input the adjacency matrix of the topology and calculate the degree and normalized results of betweenness of each node. 2) Secondly, according to the result of 1), we calculate entropy coefficient by using entropy formula 3) And then, we output the contribution matrix, according 1), 2) and the definition in Section III-B. 4) Finally, we output all the normalized value of node importance for each one. More details are shown in pseudo code.
1251
A. Multi-rooted Tree Experiment Figure 3 shows the network topology of Tree Architecture which has been marked for each node. As is shown in Figure 3, we can get the adjacency matrix and named as Btree. V2
V1
V4
V3
V7
V8
V5
V11
V10
V9
V6
V12
V13
V14
Pseudo code Begin Input: Adjacency Matrix B Output: The evaluation results of node importance of each node (1,2,...,n) 1) Compute node betweenness and normalized nodebetweenness_centrality(sparse(B)); sum (node) ; for i1 to n BCnode (i)/sum (node); // normalized results of betweenness of each node end for 2) Calculate the degree of each node for i1 to n C=B(j,:) du1=sum(C) // Calculate the degree and named as DU end for 3) Degree and betweenness results are combined into a matrix BDU for i1 to n BDU[1]DU[1] // Assign the first row of DU matrix to BDU’s first row BDU[2]DU[2] // Assign the first row of the matrix BC to BDU’s second row end for 4) Calculate entropy coefficient of both degree and betweenness Input: BDU Output: The entropy coefficients of both degree and betweenness for j1 to m //The calculation of entropy fij1sj(I,:)\R(i,:) H (fij2.*log(fij2)) end for 5) Calculate the weighted betweenness and degree and contribution matrix for i1 to n output x[i]; // Output normalized results of node importance end for end
IV. EXPERIMENT EVALUATION In this section, we demonstrate a thorough experimental evaluation of the proposed technique on Multi-rooted Tree and FiConn Architectures. The whole experiment project is implemented by Matlab 7.0 on a Windows7 Operating System with Intel Core i3 Processor 2.10GHz. In Section A, we describe and discuss the experimental results of Multi-rooted tree [7], as well as FiConn experiment and discuss in Section B.
© 2013 ACADEMY PUBLISHER
Figure 3. Multi-rooted Tree Architecture been marked Btree=[ 0 0 1 1 1 1 0 0 0 0 0 0 0 0
0 0 1 1 1 1 0 0 0 0 0 0 0 0
1 1 0 0 0 0 1 1 0 0 0 0 0 0
1 1 0 0 0 0 0 0 1 1 0 0 0 0
1 1 0 0 0 0 0 0 0 0 1 1 0 0
1 1 0 0 0 0 0 0 0 0 0 0 1 1
0 0 1 0 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0 0 0
0; 0; 0; 0; 0; 1; 0; 0; 0; 0; 0; 0; 0; 0 ];
TABLE I. DEGREE AND BETWEENNESS OF TREE Node degree betweenness V1 4.0000 0.1837 V2 4.0000 0.1837 V3 4.0000 0.1582 V4 4.0000 0.1582 V5 4.0000 0.1582 V6 4.0000 0.1582 V7 1.0000 0 V8 1.0000 0 V9 1.0000 0 V10 1.0000 0 V11 1.0000 0 V12 1.0000 0 V13 1.0000 0 V14 1.0000 0 TABLE II. WEIGHT OF DEGREE AND BETWEENNESS Attribute Degree Betweenness Coefficient 0.2004 0.7996 TABLE III. RESULT OF OUR ALGORITHM AND LITERATURE [7] Our method Literature[7] Node Node importance Node importance V1 0.0963 0.0789 V2 0.0963 0.0789 V3 0.1740 0.1772 V4 0.1740 0.1772 V5 0.1740 0.1772 V6 0.1740 0.1772 V7 0.0139 0.0167 V8 0.0139 0.0167
1252
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
V9 V10 V11 V12 V13 V14
0.0139 0.0139 0.0139 0.0139 0.0139 0.0139
TABLE V. WEIGHT OF DEGREE AND BETWEENNESS Attribute Degree Betweenness Coefficient 0.2017 0.7983
0.0167 0.0167 0.0167 0.0167 0.0167 0.0167
1) According to Table I, in the case of consideration of degree only, we can see that nodes of {V1 , V2 , V3 , V4 , V5 , V6 } have the same value of importance. In addition, from the point of view of the betweenness, the node importance of {V1 , V2 } is much higher than those of {V3 , V4 , V5 , V6 } . . 2) As shown in the Table III, we see that our proposed method have the same results with the method in literature [7]. Nodes of {V3 , V4 , V5 , V6 } get the same importance, and all of them are much higher than those of {V1 , V2 } . Compared with the literature [7], evaluation results in this paper are more accuracy. 3) For one thing, the servers are concentrated in the leaf layer. For another, the server is only connected to each switch. Thus, there comes the bandwidth bottleneck problem in switches from the layer of core and aggregation in Multi-tree Rooted Architecture. B. FiConn Experiments We next return the above experiments in FiConn. Figure 4 shows the network topology of the FiConn Architecture which has been marked for each node. B2 is the adjacency matrix of FiConn. B2 = [0 1 1 1 1 0 0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 1 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0 0 0 0 0 1
1 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 1 1 1 1 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0
0 1 0 0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0 1 0 0
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 1 1 1
0 0 0 0 0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 1 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 0 1 0 0 0 0
0; 0; 0; 1; 0; 0; 0; 0; 0; 0; 1; 0; 0; 0; 0];
V1
V2
V12
V3
V5
V4
V8
V13 V14
V11
V7
V10
Figure 4. FiConn Architecture been marked
© 2013 ACADEMY PUBLISHER
TABLE VI. Result of our algorithm and literature [7] Our method Literature[7] Node Node importance Node importance V1 0.1907 0.1972 V2 0.0581 0.0524 V3 0.0132 0.0156 V4 0.0581 0.0524 V5 0.0132 0.0156 V6 0.1907 0.1972 V7 0.0132 0.0156 V8 0.0581 0.0524 V9 0.0581 0.0524 V10 0.0132 0.0156 V11 0.1907 0.1972 V12 0.0132 0.0156 V13 0.0581 0.0524 V14 0.0132 0.0156 V15 0.0581 0.0524
1) As is demonstrated in Table VI, nodes of {V1 , V6 , V11} have higher importance in FiConn, which are distributed in each recursive unit. Our proposed method gets the same results with the method in literature [7]. Therefore, our method is effective for the assessment of node importance in FiConn Architecture. After the introduction of the information entropy theory, the accuracy of this method has been greatly improved. 2) Compared with Multi-rooted Tree Architecture，for one thing, switches are not directly connected, for another, the servers are in involved in forwarding packets. Therefore, communication overhead is relatively lower. In addition, there are no concentrated bottlenecks in FiConn. However, our algorithm can still find the important nodes in each recursive unit, which are potential bottleneck points. V. RELATED WORK
V9 V6
V15
TABLE IV. DEGREE AND BETWEENNESS Node Degree Betweenness V1 4 0.1556 V2 2 0.0889 V3 1 0 V4 2 0.0889 V5 1 0 V6 4 0.1556 V7 1 0 V8 2 0.0889 V9 2 0.0889 V10 1 0 V11 4 0.1556 V12 1 0 V13 2 0.0889 V14 1 0 V15 2 0.0889
Evaluation of these new DCN architectures is mainly limited to MapReduce or scientific computing, and some of them focus on the research on how these architectures are affected in virtualized environments. however，none
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
of them concern about the security of the data center network from the perspective of node importance. The important nodes play an important part in the network attack and defense. On one hand the defenders should particularly concern about the quality of hardware and security settings; on the other hand, taking important nodes as target, the attacker will be of greater benefit. There are many methods of node importance in complex networks and communication networks (list only a few e.g. [5-7]). Corley et al firstly proposed a method to determine which are the important nodes on the shortest path in the network [5]. The main idea is to remove the node which needs to test and then analysis the loss of transmission. That means determine the node importance by analyzing the impact of the whole networks when the node is under fail state, and then decides which one is much more important. However，it unable to assess the importance of the nodes within the whole network. Node contraction [6] based on agglomerate, which depends on the degree and location of node. It does not need to remove the nodes; it preserves the integrity of the network topology. However, it fails to determine the importance for symmetric nodes. Above all, those methods mainly only consider the importance of a single node, but ignore associated characteristics of the nodes, especially when the node is removed and the whole topology changes. The method of network correlation based on the contribution matrix [7] was used in satellite mobile communications network which based on characteristics of satellite mobile communications network. That is, in the mobile satellite communication network, once the removal of the satellite nodes or central station node, there will cause the whole network collapse. Actually, the method has well dynamic effects. This method does not need to remove the node, and take full account of the correlation between the nodes. However, it does not take into account the differences between the degree and betweenness in the real networks. That means degree and betweenness differs in the node importance, the former is static while the latter is opposite thus the accuracy of this method needs to be improved. In short, direct application of these methods would not be necessarily suitable for our DCN architecture. For the consideration of inadequate of the previous ones, in this paper, we define and solve the problem of node importance in DCN for the research of vulnerability of topology. We propose a method of contribution matrix based on information entropy. All in all, our method not only remains all advantages of [7], but also greatly improves the accuracy of experiment results. VI. CONCLUSION In this paper, we investigate the problem of topology vulnerability in Data Center Network (DCN) Architectures and propose the contribution matrix method based on information entropy theory from the perspective of the importance node of in the area of topology vulnerability, and then conduct an experimental evaluation of Multi-rooted Tree Architecture and FiConn © 2013 ACADEMY PUBLISHER
1253
Architecture, each respectively as a representative of Switch-centric Architecture and Server-centric Architecture. Experiment results further show that our proposed method indeed finds the important nodes on the whole networks. In addition, our method remains the integrity of the network topology as well as has characteristics of dynamic and objective. Compared with other methods, our proposed one has much higher accuracy. In other word, our method is generic and can be widely used in the new architectures. Furthermore，our algorithm provides the foundation for the depth research of attack and defense in DCN. Consequently, it also can be a guide for the cloud computing security [16]. ACKNOWLEDGMENT This work is supported by the National Natural Science Fund China under Grant No. 2009CB320406, the National 863 High-tech Project of China under Grant No. 2011AA01A102, the Important National Science Technology Specific Projects under Grant No. 2009ZX01039-001-002, Funds for Creative Research Groups of China (60821001) and State Key Lab of Networking and Switching Technology. Ph.D. Programs Foundation of Ministry of Education (20110005130001). REFERENCES [1] L. M. Vaquero, L. Rodero-Merino, J. Caceres, and M. Lindner, "A break in the clouds: towards a cloud definition," ACM SIGCOMM Computer Communication Review, vol. 39, pp. 50-55, 2008. [2] Shangguang W, Zheng Z, Qibo S, Hua Z, Fangchun Y. “Cloud model for service selection,” in Proc. The 30th IEEE Conference on Computer Communications Workshops on Cloud Computing, Shanghai, China. p. 666671, 2011. [3] M. Al-Fares, A. Loukissas, and A. Vahdat, "A scalable, commodity data center network architecture," in ACM SIGCOMM Computer Communication Review, vol.38, pp. 63-74, 2008. [4] Y. Zhang, A. J. Su, and G. Jiang, "Understanding data center network architectures in virtualized environments: A view from multi-tier applications," Computer Networks, vol. 55, pp. 2196-2208, 2011. [5] H. Corley and D. Y. Sha, "Most vital links and nodes in weighted networks," Operations Research Letters, vol.1, pp. 157-160, 1982 [6] W. Jun and T. Yue-jin, "Finding the most vital node by node contraction in communication networks," Communications, Circuits and Systems, 2005. [7] Y. Zhao, Z. Wang, J. Zheng, and X. Guo, "Finding most vital node by node importance contribution matrix in communication networks," Journal of Beijing University of Aeronautics and Astronautics, vol.35, pp. 1076-1079, 2009. [8] C.E. Leiserson, "Fat-trees: universal networks for hardware-efficient supercomputing," Computers, IEEE Transactions on, vol. 100, pp. 892-901, 1985. [9] A. Greenberg et al. "VL2: a scalable and flexible data center network," in ACM SIGCOMM Computer Communication Review, vol.39, pp. 51-62, 2009. [10] C. Guo et al. "Dcell: a scalable and fault-tolerant network structure for data centers," in ACM SIGCOMM Computer Communication Review, vol.38, pp.75-86, 2008.
1254
[11] C. Guo et al. "BCube: a high performance, server-centric network architecture for modular data centers," in ACM SIGCOMM Computer Communication Review, vol. 39, pp. 63-74, 2009. [12] D. Li et al. "FiConn: Using backup port for server interconnection in data centers," in proc.28th IEEE International Conference on Computer Communications (INFOCOM), pp. 2276-2285, 2009. [13] X.-R. Guo et al. "Primary studies on urban ecosystem health assessment," China Environmental Science, vol. 22, pp. 525-529, 2002. [14] Z.-H. Wang, Y.-N. Han, T. Lin, Y.-M. Xu et al. “Resource allocation algorithms in the reconfigurable network based on network centrality and topology potential. Journal of China Institute of Communications,” 33(8):10-20, 2012. [15] C.-H DING, X.-W DONG, M.-L GUI, “To evaluate the Using entropy method of network node of region correspondence net,” Computer Programming Skills & Maintenance, (S1)：pp.117-118, 2009 [16] M. Ahmed, Y. Xiang, and S. Ali, "Above the trust and security in cloud computing: a notion towards innovation," in proc.8th International Conference on Embedded and Ubiquitous Computing (EUC), pp. 723-730, 2010.
Kai Peng, PH. D candidate, State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and telecommunications. His current research interests
© 2013 ACADEMY PUBLISHER
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
include cloud computing, data center network, and security of cloud computing. Rongheng Lin received his PhD degree State Key Laboratory of Networking and Switching Technology, in 2010. He is currently a instructor at the Beijing University of Posts and Telecommunication, China. His research includes IMS apps, security of network and cloud computing. Binbin Huang, PH. D candidate, State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications. Her current research interesting include cloud computing and data center network. Hua Zou received her PhD degree in communication and electronic system from the Beijing University of Posts and Telecommunication in 2010. She is currently a professor at the Beijing University of Posts and Telecommunication, China. Her current research interests include network intelligence and services computing. Fangchun Yang received his PhD degree in communication and electronic system from the Beijing University of Posts and Telecommunication in 1990. He is currently a professor at the Beijing University of Posts and Telecommunication, China. He has published 6 books and more than 80 papers. His current research interests include network intelligence, services computing, communications software, soft switching technology, and network security. He is a fellow of the IET.
