分散式檔案系統是目前大資料(Big Data)網路系統中主要的基礎架構之一。在本篇論文中，我們提出了一套新的分散式資料儲存架構，這個架構突破了Windows 7作業環境中的系統核心，利用載入核心模式驅動程式的方式，建立起整體網路系統的工作平台。
在這架構中大致上可分為三個部分:客戶端、主控端和儲存端。在客戶端的部份我們除了設計一個可供使用者操作的介面外，同時利用Event I/O的觀念，建立了一套核心驅動程式，使其可直接與應用層溝通，並且傳遞資料；而在主控端的部分，則是負責管理所有的儲存端，並且接收客戶端連線要求，同時我們在主控端的核心驅動中，建立了一系列的資料結構，負責記錄客戶端分散檔案的邏輯儲存路徑；最後在儲存端，則是利用多台遠端電腦，透過一致性網路存取命令實作出的邏輯硬碟，提供給客戶端使用。
為了讓我們的架構有更有效的溝通，我們在系統中定義了一套網路的協定，並透過這些網路命令來控制被安裝在系統核心中的驅動裝置，以建構我們提出的分散式資料儲存架構。
最後透過我們的實作，分別進行了上傳實驗和下載實驗。在上傳實驗中，我們透過了不同檔案的資料傳輸，求得平均的上傳速度約為8.057MB/sec。而在下載實驗中，平均的下載速度則約為8.338MB/sec。

Distributed file system is one of the basic architecture in the Big Data network system. In this paper, new distributed data storage architecture is proposed, achieving a major breakthrough in the Windows 7 operating system. The kernel-mode driver is used to establish the network working platform.
The research can be divided into three parts: client, master and storage. For the client, users can operate the system with the designed interface, which uses the idea of the event I/O to communicate with a kernel driver. The master is responsible for managing storages, and receiving commands from the client. A series of data structures are also established to record the client’s file logical path. For the storage part, consistency commands make all the remote computers transformed into logical disks, which are provided to the client.
In order to have a more effective communication, a new network protocol is defined. The commands from the protocol are for controlling the kernel driver. Therefore, the proposed distributed data storage architecture can be constructed successfully.
Two experiments are made based on the distributed data storage architecture. According to the upload experiment, the average upload speed is about 8.057 MB per second. On the other hand, the average speed for downloading is about 8.338 MB per second.

第一章 簡介 1
1.1 研究動機 1
1.2 研究目的 2
1.3 論文大綱 3
第二章 背景知識與相關研究 4
2.1分散式檔案系統 4
2.2 WDF (Windows Driver Foundation)概述 7
2.2.1 WDM (Windows Driver Model) 7
2.2.2 WDF核心框架 9
2.2.3 WDF特點 10
2.2.4 KMDF (Kernel Mode Driver Framework) 11
2.3 WSK (WinSock Kernel)概述 14
2.3.1 WSK架構 14
2.3.2 WSK 物件 15
2.3.3 Socket類別 16
2.3.4 WSK中使用IRP 17
第三章 Windows核心驅動方式之分散式資料儲存架構 19
3.1 MPD Distributed Data Storage Architecture系統架構 19
3.2 MPD Distributed Data Storage Architecture軟體架構 21
3.2.1客戶端應用程式介面(Client Application Interface) 21
3.2.2客戶端核心驅動(Client Kernel Driver) 25
3.2.3主控端核心驅動(Master Kernel Driver) 30
3.2.4儲存端核心驅動(Storage Kernel Driver) 37
3.3傳輸範例 42
3.3.1客戶端上傳檔案 42
3.3.2客戶端下載檔案 43
3.4資料傳輸速度量測 44
3.5核心驅動程式設計函式 47
3.5.1 Winsock Kernel Function 47
3.5.2多執行緒程式設計 49
3.5.3事件(Event)同步機制 50
3.5.4睡眠(Sleep)機制 52
第四章 實作和實驗結果 54
4.1 實驗環境 54
4.2 數據與討論 55
4.2.1 實驗一，上傳模式 56
4.2.2 實驗二，下載模式 58
第五章 結論 61
5.1 結論 61
5.1.1 主要作法 61
5.1.2 系統的限制 61
5.1.3 實驗結果 62
5.2 未來研究方向 62
參考文獻 63

[1] 顏春煌, 2008, 資訊系統與資訊管理:破除分散式檔案系統的迷思,空大學訊 Available: http://enews.open2u.com.tw/~noupd/book_up/5199/(396)027-033.pdf
[2] 李孟哲, 2012, “Personal Storage Management Model base on Personal Cloud”,碩士論文, 中國文化大學資訊管理學系暨資訊管理研究所
[3] SUN, 1985, Design and Implementation or the Sun Network File system, Available: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.14.473
[4] OpenAFS, 2012, OpenAFS, Available: http://www.openafs.org/
[5] Alex FTP Filesystem, Available: http://www.ludd.luth.se/~kavli/alex.html
[6] Google, 2003, Google File System, Available: http://static.googleusercontent.com/external_content/untrusted_dlcp/research.google.com/en//archive/gfs-sosp2003.pdf
[7] Wikipedia, Windows Driver Foundation, Available: http://zh.wikipedia.org/wiki/Windows_Driver_Foundation
[8] Yahoo blog, 2009, 驅動程式in WDM, Available: http://tw.myblog.yahoo.com/michael-qq/article?mid=300&next=298&l=f&fid=29
[9] Ronald D. Reeves, 2011, “Windows 7 Device Driver”, 初版, p.129, 碁峯出版社
[10] 張珮, 馬勇, 董鑒源, 2011, “深入淺出Windows驅動程式開發”, 初版, p3-1, 松崗出版社
[11] PIXNET, 2008, IRP(I/O Request Packet), Available: http://graffine.pixnet.net/blog/post/13597308-irp-(i-o-request-packet)
[12] Baidu, 2012, winsock kernel, Available: http://baike.baidu.com/view/5580194.htm
[13] Chia-Ping Tsai, 2012, “Design and Implementation of Cloud Data Backup System with Load Balance Strategy“, 碩士論文, 中山大學電機工程研究所
[14] grayfox, 2009, 在內核中創建線程, Available: http://nokyo.blogbus.com/logs/33484395.html
[15] Microsoft, 2013, Core Kernel Library Support Routines, Available: http://msdn.microsoft.com/en-us/library/windows/hardware/ff542078(v=vs.85).aspx
[16] Microsoft, 2013, KeDelayExecutionThread routine, Available: http://msdn.microsoft.com/en-us/library/windows/hardware/ff551986(v=vs.85).aspx
[17] Microsoft, 2013, Windows API List, Available: http://msdn.microsoft.com/en-us/library/windows/desktop/ff818516(v=vs.85).aspx
[18] Liu, Tzong-Jye, Chung, Chun-Yan,Lee, Chia-Lin,”A High Performance and Low Cost Distributed File System”, in Software Engineering and Service Science (ICSESS), 2011 IEEE 2nd International Conference on, July 2011,page 47-50
[19] Min Xu, Junrui Zhou, Wei Zhou, Hong An, “CHMasters: A Scalable and Speed-efficient Metadata Service in Distributed File System”, 2011 12th International Conference on Parallel and Distributed Computing, Applications and Technologies, 2011, page 397-399
[20] Quan Zhang, Dan Feng, Fang Wang, “Metadata Performance Optimization in Distributed File System”, 2012 IEEE/ACIS 11th International Conference on Computer and Information Science, 2011, page 476-481
[21]H. Lamehamedi; B. Szymanski; Z. Shentu; E. Deelman, “Data replication strategies in grid environments” in Proc. Fifth International Conference on Algorithms and Architectures for Parallel Processing, ICAPP’2002, Oct. 2002, page 378 - 383.
[22] Corujo, D., Aguiar, Rui L. ; Vidal, I.; Garcia-Reinoso, J., “A Named Data Networking Flexible Framework for Management Communications”, Communications Magazine, IEEE, December 2012, page 36 – 43