System virtualization can aggregate the functionality of multiple standalone computer systems into a single hardware computer. It is significant to virtualize the computing nodes with multi-core processors in the cluster system, in order to promote the usage of the hardware while decrease the cost of the power. In the virtualized cluster system, multiple virtual machines are running on a computing node. However, it is a challenging issue to automatically balance the workload in virtual machines on each physical computing node, which is different from the traditional cluster system's load balance. In this paper, we propose a management framework for the virtualized cluster system, and present an automatic performance tuning strategy to balance the workload in the virtualized cluster system. We implement a working prototype of the management framework (VEMan) based on Xen, and test the performance of the tuning strategy on a virtualized heterogeneous cluster system. The experimental result indicates that the management framework and tuning strategy are feasible to improve the performance of the virtualized cluster system.

The virtualization technology makes it feasible that multiple guest operating systems run on a single physical machine. It is the virtual machine monitor that dynamically maps the virtual CPU of virtual machines to physical CPUs according to the scheduling strategy. The scheduling strategy in Xen schedules virtual CPUs of a virtual machines asynchronously while guarantees the proportion of the CPU time corresponding to its weight, maximizing the throughput of the system. However, this scheduling strategy may deteriorate the performance when the virtual machine is used to execute the concurrent applications such as parallel programs or multithreaded programs. In this paper, we analyze the CPU scheduling problem in the virtual machine monitor theoretically, and the result is that the asynchronous CPU scheduling strategy will waste considerable physical CPU time when the system workload is the concurrent application. Then, we present a hybrid scheduling framework for the CPU scheduling in the virtual machine monitor. There are two types of virtual machines in the system: the high-throughput type and the concurrent type. The virtual machine can be set as the concurrent type when the majority of its workload is concurrent applications in order to reduce the cost of synchronization. Otherwise, it is set as the high-throughput type as the default. Moreover, we implement the hybrid scheduling framework based on Xen, and we will give a description of our implementation in details. At last, we test the performance of the presented scheduling framework and strategy based on the multi-core platform, and the experiment result indicates that the scheduling framework and strategy is feasible to improve the performance of the virtual machine system.

Networked Virtual Design Environment(NVDE) realizes collaborative design for multiple persons in different places,focuses on aircraft design at present as its application,and is based on Networked Virtual Environment(NVE).We firstly introduce the network centric architecture of NVDE,and then explain two key technologies:implementation of design on single platform by using Windows-based High Performance Computing(HPC),and supporting collaborative design on optical network by using VLAN.Finally,we summarize the progress and future work of NVDE.

In wireless sensor networks, the data aggregation is an essential paradigm for routing, through which the multiple data from different sensors can be aggregated into a single data at intermedial nodes enroute, in order to eliminate data redundancy and achieve the goal of saving energy. Some existed medium access protocols and algorithms can effectively prolong the lifetime of the sensor network by determining when each sensor should transmit its data, and when it should sleep. In this paper, we focus on applying multiple spanning trees to organize the data aggregation, which is different from these existed single spanning tree methods. At first, the problem of constructing multiple spanning trees is transformed into a linear programming problem of the data flow network. Based on the solved optimal rate between the two adjacent sensors, the two constructing algorithms of the spanning tree are presented. Experimental results indicate that the method of multiple spanning trees can be of benefit to energy saving for wireless sensor networks, and the corresponding appropriate constructing algorithm can prolong the lifetime of the sensor network.

Advances in computer technology and network technology provide a chance that multiple distributed computational resources can be shared by Internet to solve large-scale computing problems. In this paper, we focus on computational resources, and present an incentive sharing approach for this kind of resources in the autonomous environment. Firstly, we describe the sharing scenario, in which the barter auction mechanism is adopted and the deed is used to keep the trace of trades between different resource control domains. Then we discuss the evaluation methodology for testing the proposed approach. Thirdly, the auction schema is described, and the calling strategy and the responding strategy of auction are introduced. The simulation is performed based on the synthetic workloads to evaluate the performance of the auction strategies, and experimental results indicate that the barter auction method can allow computational resource domains to provide better computing service to their users.

The virtual machine system such as Xen provides a security isolation between virtual machines (VM) running on the virtual machine monitor (VMM). With the wide application of the virtualization technology, VMM is expected to not only provide the simple isolation but also provide limited sharing between VMs in a secure manner. In this paper, we present an access control mechanism for the virtual machine system, which is based on the BLP model. We prove that the virtual machine system with the access control mechanism and an initial secure state is a secure system. In addition, we implement a prototype of the access control mechanism for the virtual machine system based on Xen.

Resources in the grid context belong to different control organizations with different interest, therefore the economic interest of each grid participant should be considered. The economic mechanism can guarantee the interest of participants in the grid with fairness and efficiency. In this paper, an economic-based resource management framework is put forward for grid computing, and then how to determine the price of resources with the economic mechanism is studied. A general equilibrium method is presented for general resources and a double auction method is proposed for special resources in the grid environment, respectively. Simulations are performed and experimental results indicate that the two methods are effective for corresponding application scenarios.

SAGE(ShanghaiGrid Adaptive Grid Engine)is a pure C++ distributed grid middleware we implemented. It's built for communications between desktop applications and cluster computing services. SAGE has been successfully set up in Intelligent Traffic Information System of ShanghaiGrid II. The article introduces the SAGE distributed parallel computing model, SAGE features, and the detail implementations. Finally, the performance evaluation is given.

Due to using the simplified interference model, IEEE 802.11 MAC protocol introduces the hidden and exposed terminal problems which significantly decrease the performance of ad hoc wireless networks. In this paper, we propose a novel Interference Graph based MAC protocol (IG-MAC) to improve the throughput of ad hoc wireless networks. The key point is to model the interference information by means of Interference Graph and send busy tone with encoded communication information to prevent the potentially interfering nodes from initiating new transmissions. Through the simulation, our protocol can solve the above two problems caused by 802.11 and improve the network performance substantially.

在网格环境中,实现动态的负载平衡在服务调度中起到了非常关键的作用。然而,网格环境中的资源的动态性决定了它难以被监控,自治性又决定了它难以被集中管理。为在网格中间件中实现服务调度的负载平衡,提供了一些具有参考价值的解决方案。介绍设计和实现调度框架的动机以及需要解决的问题,并深入探讨在上海网格中间件中设计和实现调度框架的思路,以及达到动态负载平衡的具体方案。

With the development of grid technology and Web service, service computing comes into being and is developing with time. In this paper, high performance computing is restudied based on the service-oriented architecture, which had been implemented in traditional computational grid. Firstly, according to the characteristic of high performance computing applications, a hierarchical resource management architecture is presented in combination with the service-oriented principle, which includes grid portal, global resource management level and local resource management level. Secondly, the program structure of high performance computing applications in the grid environment is analyzed and represented by the directed acyclic graph (DAG). Thirdly, a modified dynamic level scheduling algorithm is presented in accordance with the above resource management architecture and the high performance computing application model. Finally, the performance of the presented algorithm is tested with the simulation experiments, and experimental results show that the presented algorithm is suitable for the grid environment, further validate that the presented strategy for the service-oriented high performance computing is effective.

中国教育科研网格ChinaGrid是教育部在十五“211”项目支持下启动的,得到国家科技部863高性能计算重大专项支持的研究项目。项目的出发点是充分利用中国教育和科研计算机网CERNET优良的基础设施和其丰富、优质的各类资源,在中国建设一个最大、最先进、最实用的网格。首批有12所学校参加,目前已经发展到20所学校,已经初步形成包括生物信息、图像处理、计算流体力学、大学课程在线和海量信息处理在内的5大类典型应用。目前,中国教育科研网格ChinaGrid正处于第一期结束,第二期规划开始。自本期起,本刊将对ChinaGrid一期的重点研究成果进行回顾,并介绍ChinaGrid二期建设的部分设想及总体部署。

针对计算网格资源的特点以及运用经济机制进行网格资源管理所具有的灵活性及有效性,提出一种改进的基于双向拍卖机制的网格资源分配方法.首先,描述了基于双向拍卖机制的资源分配框架,整个系统由买方、卖方和计算资源经纪人组成.然后,针对网格中的CPU资源,提出一种改进的双向拍卖机制,采用统一拍卖方式,可以灵活调节交易双方的付费.进而,分析了该双向拍卖机制满足优势策略激励相容、预算平衡以及个人理性的特点,并定义了拍卖机制的效率.最后,通过实验分析了双向拍卖分配机制的效率.

与串行程序相比,并行程序调试会遇到新的问题.首先并行程序往往需要长时间运行,从而导致并行程序调试是一个尤其费时的过程;其次并行程序调试过程中,某一次调试出现的错误在下次调试的时候不一定出现,给错误跟踪带来了很大困难.本文针对这两个问题,设计和实现了一个中间件系统,在并行调试工具XMPI中使能BLCR检查点系统的.通过该中间件,在使用XMPI调试大型MPI并行程序的时候,减少调试阶段并行程序运行时间,并且可以更好跟踪并行程序错误,提高并行程序开发效率.

In this paper, we challenge the issue of resource management and scheduling in the grid context, which is compliant with WS-Resource Framework. Firstly, we focus on the high performance application, and model the large-scale scientific computing problem that can be decomposed into multiple sub-problems, which evolves to be represented by a DAG. Then, a hierarchical infrastructure for resource management in the grid context is proposed in accordance with the specifications of WS-Resource Framework. Thirdly, we discuss the scheduling issue in the presented scenario and present a modification of the DLS algorithm. At last we analyze the presented modified algorithm with simulation experiments.

In this paper, we propose a user-guided semi-automatic parallelization method, which is based on code templates corresponding to parallel programming paradigms and the concept of meta-task independent with each other. As an implementation of this method, we develop the system Metaparallel, which is based on Java language and MPICH, and the framework of Metaparallel is discussed. At last, the parallelization flow is studied with a case. In addition, we test the usability of Metaparallel by the practical engineering problem.

Up to now, there have been three un-compatible standards in service-oriented grid environment: Standard Web Service (WS), OGSI (Open Grid Service Infrastructure), WSRF (Web Service Resource Framework). In order to make use of these distinct services in a consistent way, we propose the Service Adapter solution. In this paper, following a brief description of ShanghaiGrid Core (SG-Core), we focus on the design and implementation of Service Adapter that used in the SG-Core. Service Adapter simplifies the invocation procedure of the un-compatible services, and it provides user with a transparent service access interface. It is designed in an extensible way, so new standard adapter or binding extension can be added easily.

在异构计算环境中负载平衡是一个重要问题.移动代理是一种新的分布计算模式,具有许多优势,比如移动代理能够从一台机器移动到另一台机器执行任务.该文提出了一个基于移动代理的并行计算框架,利用一个二段负载平衡策略使程序能够适应不断变化的异构计算环境.实验结果显示移动代理不仅能够用于并行计算,而且能够有效地改善负载平衡.