模糊群體決策之運算協調機制

Abstract

隨著資訊系統的資料處理量和運算量的日益增加，再加上電腦網路蓬勃發展，使得分散式計算比傳統集中式計算更受重視，各種分散式計算環境也應運而生。Computing Power Services（CPS）是一個輕量級以網路服務（Web Services）為基礎的運算能力分享環境，適合在企業內的信賴網路中執行運算工作，此架構使用Business Process Execution Language（BPEL）及Oracle的BPEL流程引擎提供可圖形化發展環境和流程控制管理。然而，在類似CPS這種即時分散式運算環境中，終會面臨電腦資源運用不當的窘境，造成部份電腦工作量過大，有些卻只處理少量工作，這種情況除了使系統各節點負載不平衡之外，還會導致運算資源浪費，為了達到較佳的運算效能與整體工作效率，「協調運算資源」是其中重要的關鍵。 本研究利用模糊群體決策方法，提供CPS架構具有服務品質（Quality of Service）的即時運算協調能力，來提升整體架構的執行效能，並使得在資源、條件不同的情形下，可保有相對穩定的運算能力。在本研究中，此方法運用在分析數位浮水印強度的濾波器組（Filter Bank）選擇，以及圖片攪亂實驗，用以測試執行效果，其結果和未使用協調機制相比，在效能及穩定度上，本研究的協調機制皆能加以改善。

As demands of data processing and computation are increasing in the booming Internet, distributed-computing environment is gaining more attention than centric-computing one. Among many proposed dictributed-computing architecture, CPS is a lightweight, Web-services-based computing sharing architecture. It is applicable to implement on trusty network in an enterprise by Web-services and Business Process Execution Language (BPEL) for providing a visualized developing environment and workflow management. However, a real-time distributed computing architecture like CPS needs to handle computing resources properly to prevent resources wasting and unbalancing. In order to have an optimized computing result, the way to coordinate computing resources will be a key factor. Therefore, a fuzzy group decision-making module is proposed in this paper to add on CPS to provide real-time computation coordination and quality of service. This module increases computing performance efficiently and shows stability in the variable sharing environment. In this research, the module has been applied to analyze the performance of digital watermark by filter bank selection and image scrambling. As the result of this study, the performance can be improved in the aspect of speedup, efficiency and process time.