Researches On Performance Enhancement Of Gridjack Surpporting Environment

As an effective means to control the complexity of software systems and to support software development and reuse, computer architecture has become a mainstream research field. Gridjack computational model is a new computer architecture as well as a new approachto development of software and hardware. On the basis of Gridjack, the research group has designed Gridjack Supporting Environment(GSE), which reduces the complexity of classic environment programming and systems development. By using abstract fusors to represent concrete entities, the GSE enables the coexistence of multi-granularity and multi-model fusors, and the loosely coupled combination of them. Meanwhile, the GSE provides development service of system-level, effective and reusable software as well as hardware. With the opportunities brought by integration of cloud computing and big data in information and communication technology(ICT) field to improve the performance of GSE, the study on how to enhance GSE performance and expand its applications is of great theoretical and practical significance to architecture-centric system development.This thesis focuses on three aspects including formal description of dynamic architecture, system security and availability to exploit formal methods and simulation to enhance GSE performance: it explores a visual hierarchical formal method of architecture description to help designers optimize and prove the system; proposes a multi-level security model based on the information flow theory and access control strategy to improve the security of the system; uses a dynamic network performance evaluation method based on multi-dimensional measurement to increase GSE availability, making the system sense the external network environment; proposes self-adaptive prediction algorithm for service node workload and dynamic cloud resource scheduling application architecture to enable on-demand Gridjack resource; advances an effective cloud resource allocation algorithm based on resource equivalence optimization to enhance the Gridjack deployment efficiency of virtual machine resource. The main research work and innovations are summarized as follows:(1) A visual hierarchical formal specification method is presented in regard to the formal description of GSE dynamic architecture. Based on morphisms and functors, we can describe network node by embedding finite state machine model into discrete event system(DES), and combine the local model and morphisms to form a composite graphical abstract structure, in accordance with transformation rules which meet consistency requirements. We can use double push-out of hierarchical graph to model the changes of network topology. Moreover, hierarchical nature of the network topology is maintained, and the computational complexity of modeling the topological transformation can be reduced with this method. This modeling method describes the distribution and concurrency features of GSE with accurate mathematical model, which can not only meet users’ visual requirements, but make visual language specification available with current tools. Furthermore, it helps verify interfaces consistency, amend irrational elements, and improve system reliability and development efficiency. The presented method also provides a new design framework and a visual simulation method for the development of distributed systems.(2) To improve the security of the GSE, a dynamic multi-level security model based on the regulation of information flow is presented. According to the objects’ security identity and subjects’ access history, the model dynamically adjusts entities’ safety signs by security transformation rules under the assumption that integrity and confidentiality identity of subjects and objects are independent. Besides, the security of the presented model was verified based on the information flow theory. As an application case, cloud security access control solution of GSE makes security access control flexible and practical by classifying isolation of different level sensitive information in cloud servers.(3) Aiming at some blind spots in GSE availability, the study comprises three aspects: GSE sensing external network environment, GSE cloud resources distribution on demand and efficient deployment of GSE virtual machines.a. A dynamic network performance evaluation method is put forward for GSE sensing external network environment, based on multi-dimensional matter-element models, indicator system of network evaluation and fuzzy analytic hierarchy process. The experiments show that the evaluation method helps network managers track and monitor network operation dynamically and quantitatively, and facilitates users to observe and control the network.b. A self-adaptive prediction algorithm combining the advantages of linear regression(LR) and BP neural network to predict workloads in cloud is proposed for GSE cloud resource on demand distribution. The experimental results show that the proposed algorithm has a relatively high accuracy in the workload prediction compared with BP neural network and LR. At the same time, the design of the resource scheduling architecture with this method is adapted for the dynamic cloud application, and provides a good reference for improving resource utilization of cloud data centres and reducing energy consumption.c. An energy-efficient resource scheduling model based on resource equivalence optimization is presented for efficient deployment of GSE virtual machines. The cloud resource scheduling problem is modeled as a constraint satisfaction problem(CSP) model, and searching process of resource allocation is optimized by pruning the equivalent resource. Experimental results show that, compared with Dynamic Power and Min PM, the proposed algorithm(Eq Power) not only improves the performance of resource allocation, but also reduces energy consumption of cloud data center.All these findings will provide more direction for further improvement of Gridjack computing performance and offer formal description method of dynamic architecture as well as security access control scheme for GSE. In addition, it will enhance the service availability of GSE, and provide theoretical support and ways of realisation for architecture-centric system development and operational entities.