Supporting Service Reliability QoS on JINI Platform

在分散式系統的環境下，系統工程師面臨了各種不同的通訊協定、作業系統、硬體設備，需要克服各種資源的異質性，使得開發分散式系統極為困難。為了簡化開發的過程，產生了中介軟體技術；中介軟體可以隱藏各式不同的作業平台、網路環境，使得系統工程師能簡單的建立大型且複雜的系統。 目前中介軟體的演化趨勢，是加入對服務品質保證的支援；而服務可靠度為分散式系統的重要議題，因此我們設計一套可靠度的保證機制。我們的系統利用元件複製技術，提升服務的可靠度，且加強設定方式，讓使用者以直覺的詞彙作設定；在元件複製技術上，利用節點上線時間分布資料，建立可靠度預測的機制，再利用可靠度資訊，分配服務複本，以加強資源效能及保證機制的準確性。 我們以JINI及RIO平台為基礎，利用可靠度預測技術，實作服務保證機制，以驗證其機制的成效。考量多樣的分散式環境，我們的可靠度保證機制可以實作於其他支援服務品質保證的中介軟體。In the distributed system, system engineer faces kinds of network protocols, operating systems and hardware, and the heterogeneity make the developing process so complex. To ease the complexity of distributed system, Middleware has appeared. Middleware can hide the different OSs and networks, and help engineers to build system large complex systems easily. The present middleware evolution trend is adding in QoS support. The service reliability is truly an important issue, so we designed and built a reliability QoS supporting mechanism. Our mechanism used replication technologies to improve service reliability, and let user specify reliability easily by intuitive phrases. About the replication technology, we built a reliability prediction mechanism based on host uptime distribution. And we use this reliability prediction mechanism to allocate service replica, which improves resource utilization and reliability QoS accuracy. We build our reliability QoS mechanism based on JINI and RIO platform, and experiment its performance. Considering the variaty of distributed environment, our reliability QoS mechanism could be implemented with any middleware that supports QoS.中文摘要 i Abstract ii Chapter 1 Introduction 1 1.1 Motivation 1 1.2 Objective 2 1.3 Organization of this thesis 3 Chapter 2 Background 4 2.1 Distributed Technology 4 2.1.1 Java RMI 5 2.1.2 Jini Middleware 6 2.2 RIO Project 8 2.2.1 RIO Architecture 8 2.2.2 RIO Features 9 2.2.3 Deploy Service in RIO 11 2.3 Replication Mechanisms 12 2.3.1 Primary-backup Replication 12 2.3.2 Active Replication 13 2.4 Related Works 14 Chapter 3 SystemOverview 16 3.1 System Goal 17 3.2 Supporting Reliability QoS 17 3.2.1 Levels of Reliability 18 3.2.2 Reliability QoS and Deploy Descriptor 19 3.2.3 Service Reliability Prediction 22 3.2.4 Assign Replica Based on Reliability Prediction 27 3.3 Enhance RIO to Support Reliable Service Deployment 29 3.3.1 Reliable Service Deploy Procedure 31 3.3.2 Periodical Reliability Maintain Procedure 32 3.3.3 Proxy Design 33 Chapter 4 Experimental Result 37 4.1 Examine the performance of Reliability QoS Mechanism 37 4.2 Examine how Reliability Prediction help Replica Allocation 41 4.2.1 Deploying Services in Networks with Different Node Types 42 4.2.2 Deploying Services in Networks with Different Maturity 43 Chapter 5 Conclusion and Future Work 44 5.1 Conclusion 44 5.2 Future Work 45 References 4