Shukuan Lin

A defect prediction model for software based on service oriented architecture using EXPERT COCOMO

Software that adopts service oriented architecture has specific features on project scope, lifecycle and technical methodology, therefore demands for specific defect management process to ensure quality. By leveraging recent best practices in projects that adopt service oriented architecture, this paper presents a defect prediction model for software based on service oriented architecture, and discusses the defect management process based on the presented model.

A Bayesian Net Based Effort Estimation Model for Service Governance Processes

This paper presents a systematic way of making effort estimation for service oriented software. By outlining major governance processes identified from service oriented best practices, an effort estimation model based on Bayesian net is introduced. Effort estimation method and project application results based on this model have been discussed thereafter.

Research on Dynamic Cache Distribution Scheduling Algorithm on Multi-Core Processors

Nowadays, multi-core processors are becoming mainstream due to their superior performance and power characteristics. With an increasing number of processor cores being integrated onto a single chip, multi-core processors require efficient organization and management of the on-chip cache resources to provide fast data accesses for concurrently running threads. Cache allocation and sharing is critical to the effective utilization of multi-core processors. Unbalanced CPU cache sharing can easily cause serious problems, such as thread starvation, priority inversion, and inadequate CPU accounting. Aiming at this situation, a new operating system scheduling algorithm which produces balanced schedules and stable system performance by reconfiguring CPU time quantum in terms of balanced cache allocation is presented. By experiment verification, the system performance stability adopting dynamic cache distribution scheduling algorithm is higher than that using conventional scheduling algorithms. Consequently, dynamic cache distribution scheduling algorithm improves the performance stability of operating system and applications effectively on multi-core architecture.

A Dynamic Load Balancing Method Based on Stability Analysis

The key issue for parallel cluster computing is the stability of load balancing method, the unstable behavior such as oscillatory action which caused by delay can intensely influence the performance of load balancing algorithm. In such a case, a linear dynamic load balancing model is proposed in this paper. By analyzing the stability of this linear model, we can not only obtain the relation between delay and load balancing gain, but also find the affect between stability and scalability of system. Based on the linear model, a more actual nonlinear model is introduced in the end. Simulation results are presented and compared with analyzing results and other load balancing methods.

Research on synthesis parameter real-time scheduling algorithm on multi-core architecture

Nowadays, multi-core processors, which have multiple processing units on a single chip, are becoming mainstream due to their superior performance and power characteristics. Meanwhile, the design of real time system on multi-core processors is being confronted with a new challenge. However, the traditional real time scheduling algorithms, such as EDF and Pfair, easily lead to many tasks missing their deadlines on multi-core processors in overload condition. Aiming at this problem, a new scheduling algorithm which adopts synthesis parameter judging priority method to extend Pfair scheduling algorithm is proposed. The experiment results show that the deadline miss rate (DMR) of synthesis parameter scheduling algorithm is lower than that of traditional scheduling methods. Consequently, the synthesis parameter scheduling algorithm improves the schedule utilization of real-time tasks and increase real time system performance on multi-core processor effectively.

Research on the Stability of Load Balancing Algorithm for Scalable Parallel Computing

In parallel cluster computing, an unscalable or unstable load balancing algorithm can intensely affect the performance of computing. To aim at this case, this paper puts forward a linear dynamic load balancing model and analyzes the stability of this linear model on the condition of existing time delay. Base on analyzing results, this paper uses a load balancing gain to control this model with the increasing system scale. In the end, a more useful nonlinear model is proposed and the simulation results are given to compare with analyzing results and other load balancing methods.

A distributed parallel QoS routing algorithm with multi-path probing

The task of Quality-of-Service(QoS) routing is to find a path in the network that satisfies certain constraints on metrics such as bandwidth, delay, delay jitter and cost. All recently-published QoS routing algorithms consider one or two metrics and most of them require the maintenance of a global network state at each node. This paper is proposed a distributed routing algorithm which require every node to maintain only its local state. The state information of intermediate nodes is collectively used to guide the routing messages along the most appropriate paths in order to maximize the success probability. It not only keeps the merits of simplicity and low link overhead in traditional QoS routing algorithm, but also can reduce resource fragments and admit more services into a network with heavy load.

A delay-based dynamic load balancing method and its stability analysis and simulation

Delay phenomenon commonly exists in most load balancing systems for parallel computing. It can cause some unstable oscillatory actions and intensely affect the performance of the load balancing system. In this case, a time delay feedback control model is presented to describe the dynamic load balancing system in parallel surrounding. By choosing proper Lyapunov-Krasovskii functionals and using Moon inequality and Schur complement lemma, the optimal control law is obtained not only for the different delay conditions but also in a scalable system scale. At last, the simulation experiments based on multithreads proved the validity of the theory and method introduced.

Risk Prediction and Measurement for Software Based on Service Oriented Architecture

The paper presents a risk prediction model for software based on service oriented architecture and discusses risk measurement based on the presented model. The risk prediction model, ECORPM, is derived from EXPERT COCOMO, providing a risk prediction baseline for software using service oriented architecture.

A Layered Service Registry Model in Government Information Resource Governance and Interchange

Government information resource catalog and interchange is becoming a very important domain in IT construction these years. Chinese government has put great attention to this area. This paper outlines specific features and issues of China¿s government information resource catalog and exchange. A government information resource and service registry model is presented accordingly, and considerations of security control, large scale catalog synchronization, and practical engineering methodology are discussed extensively.