臺灣博碩士論文加值系統

隨著網際網路的普及，與電子商務的日漸興盛，數位電子化的資料量以極快的速度持續增加。在傳統的伺服器附屬儲存架構下(Server -Attached Storage System)，僅依靠單一台伺服器提供數位資料的儲存設備與容量，已經跟不上資料量的成長速度。在此論文裡，我們以網路附屬儲存架構（Network-Attached Storage System）與分散式檔案系統（Distributed File System）為基礎，設計一套整合式的檔案儲存系統（Consolidated File System）。在此整合式檔案儲存系統中，由連結於區域網路（LAN）上的數個儲存器（Storage Box）分工合作，以組成一個具有高度容量擴充性（Scalability）的儲存架構。每一個儲存器皆具有運算能力與儲存容量，以儲存、管理其中的資料。我們也討論了數種不同儲存器的設計方式，藉由模擬實驗，以找出影響系統效能的重要因素。同時，也利用模擬實驗的結果，比較不同設計方式對系統效能的影響。

In this thesis, we propose and evaluate a new design of a Network-Attached Storage (NAS) system which is a consolidated file system in a network environment. Our objective is to build a distributed storage system with better scalability and availability. While traditional network file systems depend on a central server machine, the consolidated file system utilizes storage boxes (workstations or PCs) cooperating as peers to provide all file system services. Every storage box in the system has the capacity and the computing ability to store and control data. User transparency is achieved in our design. From the perspective of clients, the consolidated file system can be used seamlessly as a centralized storage system.
We have prototyped several key operations in this system. To provide more performance insights, we build a comprehensive system simulator and conduct extensive performance studies for the system via the simulator. This simulator has been run against a workload of a real Web server, and the system behavior with this workload is studied. Several statistics, including the count of inter-box transmission, the average response time of each request, the system throughput, the utilization of each storage box, and their relationship to the incoming request frequency and the number of storage box are explored. In addition, we disscuss various design alternatives for the consolidated file system. Several simulation results with different configurations of storage boxes are also presented and analyzed. In light of the experimental results, a consolidated file system is being implemented in full scale.

1 Introduction 1
1 .1MotivationandSummaryofOurWork..................... 1
1 .2 RelatedWorks................................... 5
1 .3 OrganizationoftheThesis............................ 5
2 Description of the Proposed System 6
2.1DescriptionofNTLayer ............................. 7
2.2 DescriptionofINTLayer............................. 9
2.3 DescriptionofCOMMLayer........................... 11
2.4 DescriptionofOPLayer ............................. 11
3 Design Consideration 12
3.1DesignConsiderationforNTLayer ....................... 1 2
3.1.1Three Design Alternatives for the placement of NT and Mapping Table 1 2
3.1.2 File-basedandDirectory-basedMappingTables............ 1 5
3.2 DesignIssuesofINTLayer............................ 1 8
3.2.1ConsiderationforCross-boxDirectory ................. 1 8
3.2.2 ConsiderationforDirectory/FileCreation ............... 1 9
4 Simulation and Experimental Studies 23
4.1SimulationApproach............................... 23
4.1.1WorkloadGeneration........................... 23
4.1.2 DescriptionoftheSimulationStateMachine.............. 24
4.2 ExperimentalStudies............................... 27
4.2.1ExperimentOne:Inter-boxtransmission................ 27
4.2.2 ExperimentTwo:Averageresponsetimeforeachoperation..... 30
4.2.3 Experiment Three : System throughput ................ 3
4.2.4 Experiment Four : System and storage box utilization ......... 33
5 Conclusions 37

[1] R. Sandberg, D. Goldberg, S. Kleiman, D.walsh, and B. Lyon, Design and Implemen-tation
of the Sun Network Filesystem, in Proceedings of the Summer 1985 USENIX,
pages 11 9─1 30, June 1 985.
[2] B. Walker, G. Popek, R. English, and et. al., The LOCUS distributed operating system,
in Proceedings of the 5th Symposium on Operating Systems Principles, pages 46─69,
October 1 983.
[3] M. Kazar., Ubik:Replicated Servers Made Easy, in Proceedings of the 2nd Workshop
on Workstation Operation Systems, pages 60─67, September 1 989.
[4] G. Popek, R. Guy, T. Page, and J. Heidemann, Replication in the Ficus Distributed
File System, in Proceedings of the Workshop on the Management of Replicated Data,
pages 5─1 0, November 1 990.
[5] B. Liskov, S. Ghemawat, R. Gruber, and et. al., Replication in the Harp File System,
in Proceedings of the 13th Symposium on Operating Systems Principles, pages 226─238,
October 1 99 .
[6] J. Kistler and M. Satyanarayanan, Disconnected Operation in the Coda File System,
in ACM Trans. on Computer Systems,10(1), pages 3─25, February 1 992.
[7] M. K. McKusick, K. Bostic, and et. al., The Design and Implementation of the 4.4BSD
Operating System, Addison-Wesley Publishing Company, Inc., 1 996.
[8] H. Howard and et al., Scale and Performance in a Distributed File System, pages 5 ─8 ,
Feb 1 998.
[9] R. Wang and T. Anderson, xFS: A Wide Area Mass Storage File System, in Fourth
Workshop on Workstation Operating Systems, pages 49─69, October 1 993.
[10] J. Wilkes, R. Golding, C. Staelin, and T. Sullivan, The HP AutoRAID Hierarchical
Storage System, in Proc. of the 15th Symp. on Operaating Systems Principles, December
1 995.
[11] J. M. N. Thomas E. Anderson, Michael D. Dahlin and et. al., Serverless Network File
Systems, in Proceedings of the 15th ACM Symposium on Operating Systems Principles,
pages 1 09─1 26, Dec 1 995.
[12] M. Rosenblum and J. K. Ousterhout, The Design and Implementation of a Log-Structured
File System, in Proceedings of the 13th ACM Symposium on Operating
Systems Principles,pages ─5, 1 99 .
[13] G. Ma and A. L. N. Reddy, An Evaluation of Storage Systems based on Network-attached
Disks, in 1998 International Conference on Parallel Processing, pages 278─285,
1 998.
[14] H. I. Gharib, P.-W. Chen, and S.-K. Chang, Naming Service in the Distributed Object
Environment for Adlets(ADDNS), in Proceedings of 4th International Workshop on
Object-Oriented Real-Time Dependable Systems, pages 76─90, Jan 1 999.
[15] G. A. Gibson, D. F. Nagle, and et. al., File Server Scaling with Network-Attached
Secure Disks, in Proceedings of ACM SIGMETRICS, pages 272─284, 1 997.
[16] R. Hurley, J. Wong, and J. Black, Performance of File Migration in a Distributed
System, in Proc. Tenth Inte’l Conf. Computer Comm., pages 685─692, November 1 990.
[17] R. Hurley, S. Yeah, J. Wong, and J. Black, Potential Benefits of File Migration in a
Heterogeneous Distributed File System, in Proc. ICCI’93: Fifth Int’l Conf. Computing
and Information, pages 1 23─1 27, May 1 993.
[18] R. Hurley and S. A. Yeap, File Migration and File Replication:A Symbiotic Relationship,
in IEEE Transactions on Parallel and Distributed Systems, pages 578─586, June 1 996.
[19] R. K. Khattar, M. S. Murphy, G. J. Tarella, and K. E. Nystrom, Introduction to Storage
Area Network, SAN, IBM International Technical Support Organization, August 1 999.
[20] S. Wilson, Managing a Fibre Channel Storage Area Network, Storage Networking
Industry Association, November 1 998.