The Analysis And Research Of Distributed Dynamic Binary Translator For Thin Client

With the help of dynamic binary translation technology, dynamic binary trans-lator (DBT) is often used for program transplantation, code optimization in the onlytime. Recently, more and more DBTs have been widely used in hardware design,system security and other felds. However, the current development of DBTs as wellas binary translation technology still has some challenges. The main challenges areas follows. First of all, due to the strong dependence with the underlying hardware,commercial DBTs always have poor scalability and compatibility; Secondly, althoughacademic DBTs make a breakthrough in scalability, their performance is often criti-cized and thus makes them can't be used in a large scale; Thirdly, the applicationsof the present dynamic binary translation technology and DBTs are almost confnedto the felds of commercial servers and personal computers, but few applications aresupported for thin clients with limited hardware resources.Based on in-depth study of the current situation of the related technology of DBT-s, this dissertation proposes a series of systematical solutions to these challenges.Compared to previous research, the study of this dissertation aims to provide morepractical, comprehensive and efective solutions. These solutions could not only behelpful for designing the DBT systems in the future, but also expand the use of DBTsin other application felds (such as thin client area).In detail, the main contributions of this paper are the following:1. We design a commonly used virtual intermediate instruction set called V-IISfor DBTs. With the help of this virtual intermediate instruction set, the DBTscan better achieve the goal of "multi-source and multi-target" without too muchloss of the performance. This V-IIS also helps reduce the complexity of DBTs'development, and save human and material resources too; 2. We design a distributed DBT framework–DistriBit for thin clients. In this Dis-triBit framework, according to the function and the computing power of theserver and thin client, we arrange the division of diferent parts. Powerful serv-er is responsible for the work of the code translation and optimization, whileresource-limited thin client is only responsible for code execution;3. We design a code cache management strategy for the thin client. Based on thecache size and code execution behavior of thin client, we design the cache man-agement strategy to adapt to this specifc situation on the server end. With theguidance of this cache management strategy, thin clients can do more complexand efcient code management.4. In a distributed virtual cloud computing environment, we study a number ofoptimization factors related to V-IIS, such as the the number of virtual registersand the number of basic blocks in a trace, and get an idea how these factorsafect the performance of DistriBit system.