Xu Zhou

University of Electronic Science and Technology of China, Chengdu, Sichuan Sheng, China

Are you Xu Zhou?

Claim your profile

Publications (10)0 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: Structured P2P network offers efficient and faulttolerant routing, object location and load balancing for upper applications. In this paper, we propose a novel scheme to improve the proximity property of structured P2P network. Basing on 2-dimension virtual network coordinates, the psychically nearby nodes are clustered into the same and adjacent regions after mapping from the network coordinates space to the identifier space of DHT. Through searching the corresponding region id using RPC calls in Chord, nodes obtain their nearby neighbors quickly in pure distributed way. Results obtained via simulation of large scale topology models denote that our scheme has lower average neighbor delay than randomly sampling approach and only incurs a modest additional searching overhead.
    Sixth International Conference on Computer and Information Technology (CIT 2006), 20-22 September 2006, Seoul, Korea; 01/2006
  • [Show abstract] [Hide abstract]
    ABSTRACT: The feature of the peer-to-peer system such as user anonymity, open nature makes that some peers are not responsible for their distributing inauthentic information. An effective solution is that set up the trust model in the p2p system. This paper presents a novel global trust model based on confirmation theory. The reputation of peers is calculated by C-F model according the bartering history, and then the peer decides to download the sharing file from which the peer has good reputation. Analyses and simulations show that this trust model can discard the malicious peer from p2p system effectively and improve the rate of successful barter greatly.
    Operating Systems Review. 01/2005; 39:56-62.
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we describe adPD, an improved parallel downloading approach for Peer-to-Peer environment. adPD assigns each server equal portion of file as large as possible, and ensures client downloading from each server without interrupt. And after faster servers finish their own work, adPD reallocates part of unfinished work of some slower servers to them. This reallocation is proportional to the ratio of server speed. By this means, adPD dynamically adjusts the proportion of a file retrieved from each server during the downloading process to match the bandwidth of each connection, and also reduces the number of block requests as many as possible. Our experiment results show that adPD is an effective parallel download scheme in Peer-to-Peer environment.
    Operating Systems Review. 01/2005; 39:63-69.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Along with wide application of e-mail nowadays, many spam e-mails flood into people's email inboxes and bring catastrophe to their study and work. This paper presents a novel anti-spam e-mail filter based-LVQ network in terms of spam e-mails which are mainly made up of several kinds commercial or political spam emails at present. Our experiment has proved that the filter based on LVQ is superior to Bayes-based and BP-based approaches in total performances apparently.
    Operating Systems Review. 01/2005; 39:34-39.
  • [Show abstract] [Hide abstract]
    ABSTRACT: This paper presents a novel distributed cooperative fault-tolerance storage mechanism (DCFS) based on file replication and XOR verification. By using XOR verification files, DCFS makes a group of files, which are distributed among different hosts cooperative, so that not only a single file in the group can use its XOR verification files to improve its own availability, but also the availability of the group can be improved greatly. Using the graph description of DCFS, we present the principle of DCFS and give a quantitative analysis to its performance. Experimental results have provided evidence of high fault-tolerance of DCFS.
    Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on; 09/2004
  • Operating Systems Review. 01/2004; 38:90-96.
  • 2004 International Conference on Computer and Information Technology (CIT 2004), 14-16 September 2004, Wuhan, China; 01/2004
  • [Show abstract] [Hide abstract]
    ABSTRACT: To improve unstructured peer-to-peer system performance, one wants to minimize the number of peers that have to be probed before the query is answered. In this paper, we propose a replication strategy with the consideration of heterogeneous resources among peers in the unstructured P2P system. In our strategy, the system copies the replication of the popular file to high capacity peers. Experimental result shows our replication strategy reduces the network traffic greatly, and improves the popular file availability effectively.
    Parallel and Distributed Computing: Applications and Technologies, 5th International Conference, PDCAT 2004, Singapore, December 8-10, 2004, Proceedings; 01/2004
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Computerized data is becoming critical to the survival of an enterprise. RAID is a popular mechanism to offer fault-tolerance storage. But RAID can not work when two more disks fail and the distance it can reach is very limited. This paper presents a novel distributed high fault-tolerance storage mechanism called DFTS. DFTS makes use of disk space of the node in distributed network to build a public storage space which users can access like local disk. File is stored in form of original file fragments and their mutual XOR verification fragments among the nodes of the network. File is available even when a part of nodes fail. DFTS realizes long-distance fault-tolerance storage. In addition, performance of the DFTS is more than an order of magnitude higher than that of single computer. The experiments in this paper show that DFTS can offer high fault-tolerance and advanced performance.
    Operating Systems Review. 01/2003; 37:19-24.
  • [Show abstract] [Hide abstract]
    ABSTRACT: High reliability is the primary requirement for messaging system. Messaging system always utilizes disk queue to temporarily store waiting message. This paper analyzes by experiments and points out that disk queue I/O is the primary performance bottleneck in messaging system. To efficiently reduce disk I/O of messaging system, we propose a novel disk queue called SmartQ which utilizes pre-assigned continuous disk space to act as disk queue to reduce file manage overhead. Lazy Gathering Write and Sequential Grouping Prefetching polices of the SmartQ not only reduce the number of write and read, but also make full use of the disk bandwidth. Experiment shows that performance of the SmartQ is more than an order of magnitude higher than that of traditional disk queue.
    Operating Systems Review. 01/2003; 37:55-60.

Publication Stats

71 Citations

Top Journals


  • 2004
    • University of Electronic Science and Technology of China
      • School of Computer Science and Engineering
      Chengdu, Sichuan Sheng, China