摘要
随着互联网的发展,人们对于网络服务质量的要求越来越高,传统网络应用客户/服务器构架很难支撑起巨大的用户数量。对等网络极大的扩充了服务的外延,挖掘边缘网络上的资源,让客户参与服务,弥补了传统客户/服务器模式的一些缺陷。但当前的p2p系统都针对专门的应用,没有形成对资源系统的管理,降低了资源的利用率。因此,8010教研室在多年分布式并行系统的研发基础上,结合对等网络本身的分布式特征,开发出了数字有机体系统,以资源管理为根本出发点,满足不同的网络应用需要。
数字有机体系统的主要任务是信息管理和任务调度。在信息管理中,如何实现资源信息的访问是研究的一个重点:为了适应大规模的不同网络应用,资源信息的访问模式必须具有通用性和扩展性;怎样提高资源信息的访问效率是设计实现当中必须考虑的重点问题。
通过对p2p系统,传统分布式系统和文件系统的研究,本文设计和实现了数字有机体资源访问系统。数字有机体系统资源访问实现了文件系统调用、函数库两种访问接口。文件系统将数字有机体中的资源按照目录树的形式进行组织和管理,方便程序员用普通文件系统调用的形式来访问资源,极大地扩充了数字有机体的应用范围。函数库接口则扩充了资源访问的模式,为不同形态的资源访问提供更大的扩展性。通过模块化和层次化的设计使资源访问系统满足各种不同网络应用的需要而且具有极强的扩展性。对于资源访问效率的提升贯穿资源访问设计实现的始终。
测试表明数字有机体资源访问系统所实现的文件系统调用接口和函数库接口达到了设计的功能要求,在读写性能方面能很好的满足一般用户的使用需求。
With the development of Internet, people pay more attention to the network service quality. The applications of the traditional client / server architecture model are very difficult to support a huge number of users. These years, the popularity of P2P technology extremely large the extension of service, let a customer play the role of a server,make up for the traditional client / server model. But all these P2P tools focus on a particular application, lacking a systematic way for managing resources, so the utility of resources was at a low level. Hence, a new generation of distributed system—DOS (Digital Organism System) has been designed to arrange the resources and storage spaces of computers in a wide area network in the 8010 research office. DOS which combines with distributed character of P2P focuses on resource management to meet the need of various network applications.
The primary tast of DOS is resource management and task dispatch. For the resource management, the most important part is resource accessing,the resource accessing system must be flexible and expansibile to meet the need of various network applications,furthermore, the efficiency of resource accessing must be taken into consideration all the time.
The document designes and realizes resource accessing system of DOS on the analyzing of P2P system and traditional distributing system. The resource accessing system of DOS has realized the file system and the function library interfaces. The file system organizes and manages the resource in the form of directory tree. It is convenient for the programmers to access the resource by the file system call, so DOS could be used in different kinds of applications. The function library interface provides another different type of resources visiting.We make use of the modularization and layering design to make the resource accessing system more flexible and expansibile and the resource accessing more efficient.
Tests show that file system call interface and the function library interface achieve the design of the functional requirements, the reading and writing performance of DOS can be good to satisfy the needs of users.
引文
[1] 朱海滨,蔡开裕,樊爱华, 等 分布式系统原理与设计. 长沙:国防科技大学出版社,1997
[2] InterMezzo http://www.inter-mezzo.org/
[3] ASF,http://www.openafs.org/
[4] RFC1094. http://www.scit.wlv.ac.uk/ jphb/comms/nfs.html.
[5] R. Sandberg. The sun network file system: Design, implementation and experience. Tech. Report, Mountain View CA: Sun Microsystems, 1987.
[6] Brian Pawlowski,Chet Juszczak, Peter Staubach, et al. NFS version 3 design and implementation. In Proceedings of the summer USENIX Conference, June 1994
[7] Napster Inc. http://www.napster.com
[8] S.M. Lui and S.H. Kwok. Interoperability of peer-to-peer fille sharing protocols. ACM SIGe- com Exchanges, August 2002. 3(3):25–33
[9] Gnutella. http://www.gnutelliums.com
[10] Freenet. http://freenet.sourceforge.net/.
[11] Frank Dabek, Jinyang Li, Emil Sit, et al. Designing a DHT for low latency and high throughput. MIT Computer Science and Artifcial Intelligence Laboratory.2004
[12] Scalable Distributed Data Structures. http://p2pfoundation.net/Distributed_Hash_Table
[13] Distributed Hash Table, Technical Definition.http://ceria.dauphine.fr/SDDS-bibliograhie.html
[14] S. Ratnasamy, P. Francis, M. Handley, et al. A scalable content addressable network. In Proceedings of ACM SIGCOMM 2001, pages 161–172
[15] I. Stoica, R. Morris, D. Karger, et al. Chord: A scalable peer-to-peer lookup service for internet applications. In Proceedings of the ACM SIGCOMM’01 Conference (SanDiego, California, August 2001).
[16] D. Malkhi, M. Naor, and D. Ratajczak. Viceroy: A scalable and dynamic emulation of the butterfly. In Proceedings of the 21st ACM Symposium on Principles of Distributed Computing(PODC ’02), 2002.
[17] A. Rowstron and P. Druschel. Pastry: Scalable, distributed object location and routing for large-scale peer-to-peer systems. Lecture Notes in Computer Science 2218, 2001. pages 329–350
[18] R. Dingledine, M.J. Freedman, and D. Molnar. The free haven project: Distributed anonymous storage service. In: Proc. of the Workshop on Design Issues in Anonymity and Unobservabil-ity, 2000. pages 67–95
[19] J. Kubiatowicz, D. Bindel, Y. Chen, et al. Oceanstore: An architecture for global-scale persistent storage. Proceedings of ACM ASPLOS, November 2000.
[20] Ben Y.Zhao, John Kubiatowicz, Anthony D.Joseph. Tapestry: An Infrastructure for Fault-tolerant Wide-area Location and Routing. Computer Science Division University of California,Berkeley.2001
[21] A. Adya, W. J. Bolosky, and etc M. Castro. Farsite: Federated, available, and reliable storage for an incompletely trusted environment. In 5th Symposium on Operating Systems Design and Implementation (OSDI 2002), Boston, MA, December 2002.
[22] X. Fei, Y. Guangwen, and J. Dapeng. Design of distributed storage system on peer-to-peer structure. Journal of Software(Chinese), October 2004. 15(2):268–277
[23] H. Hua, G. Chaoyang, D. Yafei, et al. Global file store: A massive storage system on the internet - concept and design. ICCNMC01 proceedings, IEEE Computer Society Press, October 2001. pages 165–175
[24] 王典荫,刘心松 下一代计算机系统—数字有机体 西部广播电视[J] 2005,Vol 1,p 4-6.
[25] Ragib Hasan,Zahid Anwar, William Yurcik, et al. A Survey of Peer-to-Peer Storage Techniques for Distributed File Systems. Proceeding of the International Conference on Information Technology:Coding and Computing(ITCC’05),2005
[26] 严功军.分布式并行文件系统 DPFS 的 SSI 技术:[硕士学位论文],成都:电子科技大学,2004
[27] Hirade, Motoichi, Tanaka, Eiichi. Organization of a file system using class name expressions based on a B-tree. Systems and Computers, Japan 27 1,Jan 1996
[28] Sinha,Pradeep Kumar,Maekawa,Mamoru. Flexible user-definable performance of name resolution operations in distributed file systems. Proceedings of the 5th IEEE Symposium on Parallel and Distributed Processing Dec 1-4 1993
[29] 周伦,刘心松,丘元杰等。一种基于广域网的资源管理模型。ACA’06,Vol33-7,pg400~402
[30] Antony Rowstron, Peter Druschel. Pastry: Scalable, decentralized object location and routing for large-scale peer-to-peer systems. the 18th IFIP/ACM International Conference on Distributed Systems Plat-forms (Middleware 2001). Heidelberg, Germany, November 2001.
[31] 王强,刘心松,邱元杰. 一种广域网上的副本创建策略. ACA’06. Vol. 33, pg 390~392.
[32] DANIEL .P.BOVET & MARCO CESATI.深入理解 LINUX 内核(陈莉君等译).北京:中国电力出版社,2004
[33] 毛德操,胡希明. Linux 内核源代码情景分析.浙江:浙江大学出版社,2001
[34] William Stallings. operating systems internals and design principles.Prentice Hall.1998.
[35] Jiang Lan, Xiaotao Liu, Prashant Shenoy ,et al. Consistency Maintenance In Peer-to-Peer File Sharing Networks. Proceedings of the Third IEEE Workshop on Internet Applications(WIAPP’03), 2003
[36] Maurice J.Bash..UNIX 操作系统设计(陈葆钰,王旭,柳纯录,冯雪山译).北京:机械工业出版社,2000
[37] 李庆虎.基于 p2p 构架的网格文件系统研究:[博士学位论文],北京:清华大学,2004