Current Issue

• 2013, No. 12　Published:25 December 2013 Last issue    Next issue
  Columns in this issue: 论文

• 论文

  Research on GPGPU performance models 

  WANG Feng,DU Yunfei,CHEN Juan

  2013, 35(12): 1-7. doi:

  Abstract ( 146 )   PDF (666KB) ( 308 )   　　

  The emerging and the development of the GPGPU afford the massive computation power to the parallel applications. How to use this computation power efficiently relies on the optimization of the applications. The performance models play an important roles on the targeting the performance bottleneck, guiding the optimization strategies, load balancing with other devices, etc. The stateofart of the GPGPU performance model is described, categorized, and analyzed in details. The models are divided into statisticbased curvefitting model and analysis model. The latter one can further be divided into metric model, computation and memory parallel aware model and componentbased quantitative model. The pros and cons of all the models are analyzed, and an interpolation performance model based on statistics is implemented. Finally the unsolved problems and the future challenges are presented.

  Optimization of small file access on
  object based parallel file system  

  ZHOU Enqiang，DONG Yong，ZHANG Wei，LU Yutong

  2013, 35(12): 8-13. doi:

  Abstract ( 191 )   PDF (687KB) ( 414 )   　　

  Object based parallel file systems are widely used in current supercomputers for its excellent performance of large file at the expense of its lower efficiency on small file access. Through analysis of the overhead of small file access on typical object based parallel file system, a serverside small file aware cache is proposed in order to reduce file access latency by aggressively prefetching a set of related files object and improving small file hit rate. This method is implemented and evaluated in Lustrefile system which is a popular parallel file system for HPC system. The experiments in real system show that it can improve the performance of small file access dramatically, especially with the reduction of up to 88% of total cost in scenario of file archiving operation.

  Discussing for supercomputing center power system         

  SONG Fei，YAO Xinan，HU Shiping

  2013, 35(12): 14-19. doi:

  Abstract ( 119 )   PDF (744KB) ( 254 )   　　

  With the expansion of the scale of high performance computer system,

  power consumption has become one bottleneck of its domain development. The key and the core

  of building green supercomputing center is energy efficient. Firstly, power system structure

  and its optimized project are described. Secondly, energy efficient policies are classified

  and described respectively. Moreover, synthetic energy management and scheduling for

  supercomputing center is conceived. Finally, we summarize and discuss future research

  directions.

  Research on desktop virtualization technology based on SPICE protocol and its improvement solutions            

  XU Hao,LAN Yuqing

  2013, 35(12): 20-25. doi:

  Abstract ( 234 )   PDF (543KB) ( 277 )   　　

  With the

  continuous development of cloud computing, the virtual desktop solution becomes increasingly

  mature and the desktop virtualization becomes a typical cloud computing application. Virtual

  desktop technology can effectively solve those problems when using traditional personal

  computers and also reduce maintenance costs, which draws industries' attentions. As an open

  source desktop delivery protocol, SPICE has its own unique architecture and characteristics.

  Using SPICE protocol, virtual desktops can be deployed between servers and remote terminal

  clients with good quality of user experiences. However, SPICE protocol has its own

  drawbacks, and it is not suitable for lowbandwidth networks. Virtual desktop technology is

  introduced and the theories of virtual desktop delivery protocols as well as factors that

  affect virtual desktop performances are analyzed. Based on the analysis, the architecture of

  SPICE virtual desktop solution and its core technology are further analyzed. Finally the

  drawbacks of SPICE protocol are pointed out and some improvements methods without affecting

  QoE are proposed, which will play a guide role for the development of virtual desktop

  technology.

  Research on comprehensive visualization method
  of molecular structure and gene sequence data  

  SONG Chenglong,ZOU Chen,WANG Wenke,LI Si

  2013, 35(12): 26-33. doi:

  Abstract ( 111 )   PDF (1060KB) ( 257 )   　　

  With the rise and the rapid development of system

  biology, in order to explore the relationship between different levels of Bioinformatics

  relevance, the need of comprehensive visualization of biological scientific computing data

  at different levels is increasingly urgent. A method of constructing an integration

  framework to realize the comprehensive visualization of molecular structure data and genomic

  sequence data is proposed. The visualization integration framework can achieve the unified

  management and unified GUI for different levels of visualization tools. By defining the

  method of describing the associated data, the framework realizes the integration of exiting

  molecular structure data and genomic sequence data and effectively supports the development

  of integrated function for visual molecular structure data and genomic sequence data. The

  method is used to develop a comprehensive visualization prototype system of molecular

  structure data and gene sequence data, realizing comprehensive visualization functions such

  as the synchronous display of both molecular structure data and genomic sequence data, and

  the display of the gene fragments and the local space structure of related molecular, and so

  on. Preliminary application and analysis show that the comprehensive visualization method is

  proposed with good scalability. The realized comprehensive visualization functions have good

  application prospects for studying the incidence relationship between molecular structure

  data and genomic sequence data.

  A dynamic worm propagation model
  based on social network  

  ZHAO Ying，YI Pingke

  2013, 35(12): 34-38. doi:

  Abstract ( 124 )   PDF (817KB) ( 363 )   　　

  In recent years, social networking worms

  propagating through users’ interactions appear frequently. These worms trick users to click

  worm links by exploiting social engineering techniques, and it poses a serious threat to the

  Internet. A dynamic worm propagation model based on social network is proposed, and the worm

  propagation process between network nodes is drawn out. Validity of the model is verified by

  the simulation, and the influence of the worm propagation is analyzed by the number of

  online users, the probability of users click worm links, the release time of immune patches

  and the quantity of users installed patches.

  A quick redirection strategy based on parent node
  backup mechanism in mobile ALM environment             

  CUI Jianqun1,YE Yongjia1,GAO Kuan1,FAN Jing2,WU Libing2

  2013, 35(12): 39-44. doi:

  Abstract ( 92 )   PDF (623KB) ( 192 )   　　

  To address solving

  the high dynamics problem of mobile Application Layer Multicast (ALM) nodes, a multicast

  system’s failure recovery strategy in the mobile environment is proposed. This strategy can

  achieve the goal of quick redirection by using active failure recovery method and the

  proposed redirection algorithm based on parent node backup. Besides, we propose to use the

  similarity of multicast tree to measure the impact of the redirection algorithm on the

  change of multicast system’s topology, thus evaluating the quick redirection strategy.

  Simulation experiment is given to show that the proposed quick redirection strategy has low

  delay and small influence on the multicast system's topology..

  RAS:Rate adaptation in IEEE 802.11 with receiver’s SNR       

  DENG Xiaoheng1,LI Xu1,LIU Qiang1,CHEN Zhigang2

  2013, 35(12): 45-51. doi:

  Abstract ( 121 )   PDF (886KB) ( 206 )   　　

  Rate Adaptation

  (RA) is designed to improve the performance of multirate feature in IEEE 802.11 PHY. One

  of the key issues in RA is to differentiate the channel error from linklayer collision,

  which is overlooked by many RA algorithms. Other than some RA algorithms that identify

  collisions at the transmitter by chances, we use receiver’s SNR to reflect the channel

  condition and mitigate the effect of collisions. A RA algorithm that utilizes ACK feedback

  mechanism is proposed to inform transmitter of receiver’s SNR so as to adaptively select

  the transmission rate. Simulation results show that our proposal mitigates the effect of

  collisions and performs well in respect of throughput and delay.

  Analysis of TCP performance in
  heterogeneous networks          

  YIN Qipeng，WU Chunqing，YU Wanrong，ZHAO Baokang，MA Yanpeng

  2013, 35(12): 52-57. doi:

  Abstract ( 93 )   PDF (1683KB) ( 276 )   　　

  At present, wired networks, wireless networks, satellite networks and high-speed

  networks all can carry the data payload, and the main features of heterogeneous networks are

  high BER (Bit Error Rate), long delay, asymmetry of bandwidth in some links, and so on.

  Since TCP is designed for normal wired networks so as not to work very well in heterogeneous

  networks. OPNET is used to simulate heterogeneous networks, and obtain the regular of how

  the BER, delay and asymmetry affect the TCP performance.  The obtained regular can provide a

  reference for experiments on optimization of TCP performance in challenged networks.

  A fair scheduling policy based on delay margin in wireless mesh networks             

  FU Qi1,2,CHEN Zhigang1,JIANG Yunxia2,YIN Fengyu2,LI Runqiu2

  2013, 35(12): 58-65. doi:

  Abstract ( 93 )   PDF (994KB) ( 238 )   　　

  In IEEE802.11ebased wireless mesh networks, the

  parameters (eg., contention windows, transmission opportunity) used in 802.11e's mechanism

  is statically assigned in order to guarantee the quality of service, which is more suitable

  to the communication nodes in one hop transmission with environment and not to the wireless

  multihop environment. In view of the problem that data flows with long delay bound may

  consume more channel time than data flows with short one so as to cause packet loss and

  networks performance degradation, a fair share scheme based on the residual packet delay

  bound, which is equally distributed among data packet, is proposed to alleviate the

  retransmission number and promote the fairness between data flows. Simulation results show

  that the proposed scheme is more effective than the original 802.11e protocol.

  An improved MMAC based MAC protocol for cognitive radio network              

  WANG Hui,GAO Ren

  2013, 35(12): 66-75. doi:

  Abstract ( 116 )   PDF (1178KB) ( 271 )   　　

  The cognitive radio network is proposed in the face of scarce spectrum resources

  and a variety of heterogeneous wireless access techniques. Media Access Control (MAC)

  protocol decides how the nodes in the networks receive and send data by shared wireless

  channels. Whether the MAC protocol can use the limited channel effectively or not plays a

  very important role in the performance of the cognitive radio network. Focusing on designing

  and implementing an excellent MAC protocol in cognitive radio networks, CRMMAC is

  proposed, which is an improved MAC protocol based on MMAC. In CRMMAC, there is a dedicated

  control channel. The network conditions, channel usage model, protocol frame structure and

  channel access strategy are presented. The simulation result shows that the proposed

  protocol has good performance.

  Detecting overlapping communities in complex networks    

  ZHANG Zhenyu,ZHANG Zhen,YANG Wenzhong,WU Xiaohong

  2013, 35(12): 76-83. doi:

  Abstract ( 128 )   PDF (1082KB) ( 297 )   　　

  Community detection is the key problem in studying complex network structure. An

  overlapping community detection algorithm SAEC based on edges is proposed based on the

  analysis of those existing means. The SAEC algorithm regards the community as a set of

  edges, defines the edge similarity and obtains the transition probability matrix. Using

  spectral clustering method, the number of communities is automatically determined by

  calculating the eigenvalues and eigenvectors of the transition probability matrix. Then, the

  overlapping community detection is completed by Kmeans algorithm. The effectiveness of the

  algorithm is verified by randomly generated test networks and real networks.

  Spectrum management based on random
  graph for cognitive radio Ad Hoc networks           

  ZHOU Jian1,2,3,YU Yong hong1,SUN Li yan1

  2013, 35(12): 84-89. doi:

  Abstract ( 89 )   PDF (1018KB) ( 241 )   　　

  Base station is not supplied to supported spectrum management for cognitive radio networks Ad Hoc, so communication among nodes may be failed because there is not a shared frequency between communication nodes. In order to deal with this question, a scheme of spectrum management based on random graph theory is proposed. Users select part of spectrums from the spare spectrum set whose number is computed by the parameter including spectrums size, network scale, link connection probability before joining in network. In communications the communication efficiency is not reduced even if every user’s spectrum resource is limited. In experiments the scheme guarantees the high probability of connective for each node having few frequencies; and the path of communication does not increase due to small spectrum scale. Therefore this scheme is suitalble to the surrounding that CR Ad Hoc network can be deployed quickly and flexibly.

  Research of IS-IS interoperability testing based on colored Petri nets           

  ZHAO Yulan,JI Chao,JIANG Fengxian,ZHU Yangyang,LEI Liting，ZHANG Zhibin

  2013, 35(12): 90-95. doi:

  Abstract ( 121 )   PDF (1058KB) ( 231 )   　　

  Internet has become an undivided part in modern life, therefore an effective and stable protocol becomes more and more important. The paper studies the interoperability testing of ISIS protocol, which is a major interior gateway protocol. Firstly, a hierarchical CPN model of the interoperability testing of ISIS protocol is established. Secondly, the full state space is automatically generated by the CPN tools, and test sequences and test cases are generated according to the test purpose. Finally, a test platform is built in Windows XP and a routing simulation software, named GNS 3, is used to imitate Cisco 7200, Cisco 3640, etc.. The obtained test cases are used to do the ISIS interoperability testing, and the test reports are given after the test results are analyzed.

  Research of AODV routing protocol based on
  link available time in the tactical MANET            

  ZHOU Jiao，TIAN Jie，DAI Chencheng，LI Tingjun

  2013, 35(12): 96-101. doi:

  Abstract ( 107 )   PDF (659KB) ( 235 )   　　

  Ondemand distance vector routing protocol AODV is one of the research hotspots in Ad hoc Networks. To ensure stable data transmission, a link available time algorithm of AODV is proposed, which increases a parameter of link available time in the route request packet and a parameter of routing path available time in the route reply packet. In the process of routing maintenance, intermediate nodes start the repair process before the link really fails. Compared with the traditional algorithm, the results obtained in NS2 simulation platform show that the improved AODV protocol enhances the reliability of the routing, improves packet delivery ratio and reduces the average endtoend transmission delay.

  Dynamic load balancing scenario based on
  multiobjective genetic algorithm             

  JIN Shan，LI Qiu

  2013, 35(12): 102-106. doi:

  Abstract ( 111 )   PDF (522KB) ( 289 )   　　

  The resource load balancing problem of the largescale and heterogeneous network is studied. Firstly, the problem is modeled and analyzed theoretically, and an objective function which satisfies the host and network constraints, is designed. Secondly, a multiobjective minimum spanning tree problem based on the model is researched, and a multiobjective genetic algorithm is devised accordingly. At last, a dynamic load balancing scenario named DSMG is proposed based on the algorithm. The simulation results show that the DSMG can balance the load effectively between the lightload nodes and the overload ones by selecting the appropriate parameters. Besides, both the resource consumption rate and the quantity proportion of the bottleneck nodes perform quite well compared with other dynamic load balancing scenario. 

  An optimization algorithm for
  largescale Internet recommender system  

  JIANG Peng,XU Feng,ZHOU Wenhuan

  2013, 35(12): 107-113. doi:

  Abstract ( 118 )   PDF (750KB) ( 293 )   　　

  Recommender system is one of the key techniques in internet applications. The system analyzes user’s behavior, and recommends products initiative to replace the passive acceptance of user requests. The recommender system can improve not only the user experience but also the user’s desire to buy something. Collaborative filtering algorithm is widely used in the recommender system. In a largescale network, traditional collaborative filtering algorithms have extreme sparseness problem, and thus being inefficient. A collaborative filtering algorithm is proposed, which is designed by the largescale network segmentation rules. The algorithm uses the idea of divide and conquer algorithm, and decomposes problems into subproblems to solve, hence reaching the optimization of algorithm.

  An adaptive message pushing strategy for mobile devices         

  LIU Yongling,LIU Wu,GUO Kehua

  2013, 35(12): 114-119. doi:

  Abstract ( 126 )   PDF (590KB) ( 230 )   　　

  Considering device diversity, power consumption, network flow consumption and message realtime in mobile devices, an adaptive message pushing strategy in mobile network is proposed. In the strategy, the running parameters of mobile device are obtained to determine the message pushing mode, and then the dispatching module will automatically convert the approach between the pollingbased and the connectionbased communication. Experiment results show the strategy can solve the terminal diversity problem, ensure the message realtime and save the power and network flow consumption in comparison to some traditional message pushing strategies.  

  A dynamic updating method of user experience time of Web services           

  JU Xinglong1,DUAN Youxiang1,MA Cheng1,XIU Nanhai1,JU Haiyan2

  2013, 35(12): 120-125. doi:

  Abstract ( 92 )   PDF (644KB) ( 226 )   　　

  Web services can not be used without the QoS information. The values of Web services’ QoS attributes are dynamic, such as user experience time. With respect to the static QoS attribute values, the dynamic QoS attribute values can reflect the actual quality status better. A new dynamic updating method, called HDBQoSDPM, of user experience time of Web services is proposed. Classic Web services model does not include the QoS information. In order to verify the correctness of the method, a Web services model, named QoSBWSA, is proposed. In the QoSBWSA model, QoS information processing function is included. The experiment shows that the proposed updating method is feasible and correct.

  Digital watermarking encryption
  algorithm based on chaotic system       

  WU Jing,JING Fengxuan,QI Fumin

  2013, 35(12): 126-133. doi:

  Abstract ( 175 )   PDF (1063KB) ( 333 )   　　

  Aiming at the issues that robustness and security of most image watermarking are not strong，a digital watermarking algorithm based on multifactor encryption and Contourlet transform is proposed. Using the good pseudo randomness and mathematical statistical characteristics and the advantages of simple and quickly generated sequences of Chebyshev map, three chaotic sequences are produced. Two of them are used to encrypt the watermarking image, the left one is used as watermarking embedded keys; The Contourlet Transform is used to extract the subband coefficients of the original carrier and the encrypted watermark, then the coefficients of the watermark is embedded into the coefficients of the original carrier, thus accomplishing the watermark embedding process. The method is implemented and experiments are performed to compare the proposed method with other methods in literatures. The result shows that the algorithm owns better effect on robustness, security and invisibility.