臺灣博碩士論文加值系統

在分散式的運算環境裡，服務搜尋系統在傳遞客戶端要求執行的服務中扮演著關鍵的角色。服務的合成機制更進一步讓服務搜尋系統得以結合既有的服務，提供更多樣化的服務。然而在一個資源受限的環境裡，如嵌入式系統，目前具有服務合成機制的服務搜尋系統就顯得過於龐大而無法使用在這類環境中。在本篇論文中，我們提出了一個服務執行的機制，使得任何可在伺服端上執行的script或程式，都能夠被視為一個服務，不管此script或程式使用何種程式語言或應用程式介面(API)開發。我們利用有向圖來表示在服務搜尋系統伺服端中不同服務，並且說明如何藉由在有向圖中找尋一條路徑來達到服務合成的目的。我們選擇服務定位協定(Service Location Protocol, SLP)來實作我們所提出的方法。實驗結果顯示在不同的測試環境下，服務執行以及服務合成的效能都在可接受的範圍。我們提出的服務執行及服務合成機制能和服務搜尋系統結合用於服務的遠端搜尋及控制，相當適合應用在普遍運算中。

In the distributed computing environment, the service discovery system plays a critical role to deliver services invoked by clients. With a service composition mechanism, the service discovery system can provide variety of services by combining the inherent services. However, for a resource-restricted environment such as an embedded system, it is impossible to put the service discovery system with service composition ability in the environment. In this thesis, we propose a service invocation mechanism in which any executable scripts or programs could be defined as a service no matter how the script or program is implemented by a specified programming language or APIs. We uses a directed graph to represent the services in a server of the service discovery system, and demonstrates how the service composition can be achieved by finding a path in the directed graph. We choose Service Location Protocol (SLP) to implement the proposed mechanisms. Experimental results show that the performance of service invocation and service composition under different testing environments is acceptable. The proposed service invocation and service composition mechanism can be integrated with a service discovery system to discover and control services remotely. It is very suitable to apply them to the pervasive computing.

Table of Contents
1. Introduction……………………………………………………………………………………1
2. Related Work……………………………………………………………………………………3
3. Service Invocation……………………………………………………………………………6
3.1 Prototype………………………………………………………………………………………6
3.2 Features ………………………………………………………………………………………7
3.3 Application: E-Classroom …………………………………………………………………8
4. Service Composition …………………………………………………………………………9
4.1 Service Representation……………………………………………………………………10
4.2 Generating a Service Invocation List Dynamically…………………………………13
4.3 Execute Services from Service Invocation List ……………………………………15
5. Implementation ………………………………………………………………………………17
5.1 Implementing Service Invocation ………………………………………………………17
5.1.1 exesrvs command in slptool……………………………………………………………17
5.1.2 New SLP Messages (SLPSrvExe, SLPSrvExeRply) ……………………………………17
5.1.3 SLPExeSrvs(): New SLP API for Service Execution ………………………………18
5.1.4 Additional Functions in slpd…………………………………………………………19
5.2 Implementing Service Composition………………………………………………………19
5.2.1 Generating Adjacency Lists in slpd…………………………………………………19
5.2.2 Modification of exesrvs command in slptool………………………………………20
5.2.3 Getting and Processing a Service Invocation List………………………………20
5.2.4 Generates a Service Invocation List Dynamically ………………………………21
6. Evaluation ……………………………………………………………………………………24
6.1 The Performance of Service Composition………………………………………………25
6.2 The Time Needed to Execute Services …………………………………………………27
6.3 The Performance of Service Invocation ………………………………………………29
7. Conclusions and Future Work………………………………………………………………31
References…………………………………………………………………………………………33
List of Figures
Figure 4.1: The directed graph represents the services in Table 4.1 ……………11
Figure 5.1: The format of SLPSrvExe message ……………………………………………18
Figure 5.2: The adjacent lists for Server A in Figure 4.1 …………………………20
Figure 5.3: The execution flow of SLPDSrvComp() ………………………………………22
Figure 5.4: The format of SLPPathRqst message …………………………………………22
Figure 5.5: The format of SLPPathRply message …………………………………………23
Figure 6.1: The generating time of service invocation list for different test cases ………………………………………………………………………………………………25
Figure 6.2: The service execution time for different test cases …………………27
Figure 6.3: The service invocation time for different test cases…………………29
List of Tables
Table 4.1: An example of service representation ………………………………………10
Table 6.1: The summary of experiment setup………………………………………………24
Table 6.2: The test cases for experiments ………………………………………………24
Table 6.3: The generating time of service invocation list …………………………26
Table 6.4: The service execution time for different number of services…………28
Table 6.5: The time to complete a service invocation request………………………30

[1] Universal Plug and Play Forum, http://upnp.org/[2] Yaron Y. Goland, Ting Cai, Paul Leach, Ye Gu, and Shivaun Albright, “Simple Service Discovery Protocol/1.0, Operating without an Arbiter”, http://upnp.org/download/draft_cai_ssdp_v1_03.txt, October 1999. [3] E. Guttman, C. Perkins, J. Veizades, and M. Day, “rfc2608: Service Location Protocol, Version 2”, http://www.ietf.org/rfc/rfc2608.txt, June 1999. [4] J. Kempf and E. Guttman, “rfc2614: An API for Service Location”, http://www.ietf.org/rfc/rfc2614.txt, June 1999. [5] Zhuoqing Morley Mao, Eric A. Brewer, and Randy H. Katz, “Fault-Tolerant, Scalable, Wide-Area Internet Service Composition”, http://www.cs.berkeley.edu/~zmao/Papers/techreport.ps.gz, January 2001. [6] The Ninja Project, http://ninja.ca.berkeley.edu/index.html[7] Jini[tm] Network Technology, http://www/sun.com/software/jini/index.html