Design and implementation of two-level metadata server in small-scale cluster file system

The reliability and high performance of metadata service is crucial to the store architecture. A novel design of a two-level metadata server file system （TTMFS） is presented, which behaves high reliability and performance. The merits both centralized management and distributed management are considered simultaneously in our design. In this file system, the advanced-metadata server is responsible for manage directory metadata and the whole namespace. The double-metadata server is responsible for maintaining file metadata. And this paper uses the Markov return model to analyze the reliability of the two-level metadata server. The experiment data indicates that the design can provide high throughput.     

A distributed service composition method based on service association

In this paper, a Web service composition architecture based on structured P2P network is proposed. Semantics is used to achieve service accurately matching and user personality customization. Through Web service virtual mapping（WVM） association the fast computing of distributed service composition based on the service function is also implemented. The Web service composition architecture and distributed service composition algorithm proposed in this paper solve a series of existent problems in ser- vice discovery and composition in distributed environment, and provide a service composition result meeting user personality requirement. At the same time, they improve the efficiency of service composition calculation.     

An Efficient Framework to Utilize Grover Search

This paper proposes an efficient framework to utilize quantum search practically.To the best of our knowledge,this is the first paper to show a concrete usage of quantum search in general programming.In our framework,we can utilize a quantum computer as a coprocessor to speed-up some parts of a program that runs on a classical computer.To do so,we propose several new ideas and techniques,such as a practical method to design a large quantum circuits for search problems and an efficient quantum comparator.     

Process View of Reflection Mechanism for Reuse Software Architecture

Reflection mechanism for reuse software architecture （RMRSA） divides a software architecture into base-level architecture and meta-level architecture logically. Base-level architecture is the ordinary architecture; meta-level represents and manipulates the reusable meta-information of base-level architecture explicitly. Through reflection, the modification of meta-level architecture will result in the modification of the architecture in base-level. Then we can gain a new base-level architecture design. In this paper, we use π-calculus to define the constituents and their interaction processes of RMRSA, by these definition, we specify the business function in base-level at runtime, and illustrate the reflection mechanism between the base-level architecture and meta-level architecture.     

Constrained Bézier curves'' best multi-degree reduction in the L2-norm

In computer aided geometric design, the degree reduction of the parameter curve is a key technique in data exchange and data compression. The various existing methods of degree reduction cannot decide whether the degree reduction curve satisfying the given tolerance exists beforehand, cannot give approximation of the best multi-degree reduction, or cannot provide explicit expression and error formula of the degree reduction curve. In this paper, we propose an entirely new method, which can hurdle the above flaws completely. For a given Bézier curve of degree n, we can easily decide whether a Bézier reduction curve of degree m exists, which has equal derivatives with the given curve up to (r-1)-th and (s-1)-th orders (r,s≤m) respectively at the endpoints, so the approximating error is less than the given tolerance ε in the L2-norm. If the curve exists, the explicit expression can be given.     

Research on Software Architecture of the Extensible Services Router

With the rapid development of network technology,it is increasingly important to dynamically upgrade router‘s software.In this paper,the authors present new software architecture to be applied to extensible services router.The novel feature of the architecture is the ability to dynamically load and configure extensible components at run time.Each component implements one function such as packets classification,packets scheduling and routing protocol.We design a simple configuration language(ESRCL)to configure and manage the router,The architecture can be configured according to the packet flows,At Present,a prototype system based on highly efficient router operating system(HEROS) has been finshed.     

应用SPOM方法辅助获取与分析面向服务需求

In order to design and provide good services,it is necessary for students to describe and understand the customer requirements for services rapidly,to design and plan the services and their behavior,and to schedule service processes effectively.This paper proposes a new service-oriented requirement elicitation and analysis method based on answer set semantic,sketches requirement Meta-model and correlative proofs and algorithms,which can help students represent requirement more declarative and precise.Based on the Meta-model,Subject-Predicate-Object requirement method and specification is presented,by which students can constitute a smooth,small-step,incremental and iterative development process cut in by forms.By applying this method to teach and help students develop service-oriented application,it is clear that it can help students eliminate the miscommunication between developers and users,assure the correctness of the artifacts and make the users active in development cycle.     

Solid Freeforming and Combinatorial Research

A view of manufacturing processes is presented in which five distinct categories are defined as casting, deformation, machining, joining, and solid freeforming. Solid freeforming is essentially biomimetic and shares problems of morphogenesis with natural processes. Our team in University of London has been exploring three mechanisms of solid freeforming. In dry powder deposition and direct ink-jet printing, the emphasis has turned to the problem of delivering a complex shape in which the three dimensional spatial arrangement of composition is delivered from the design file. In extrusion freeforming, the aim is to control microstructure at hierarchical levels also from the design file. The quest for 3-D functional gradients is satisfied by acoustic and ultrasonic dispensing and mixing of powders so that each layer can be patterned. These methods could be extended to deliver the complex patterns demanded by left-handed microwave metamaterials. Dry powder deposition and direct ink-jet printing are turning towards combinatorial methods in which multiple sample libraries are used to accelerate discovery. In turn, this paves the way for ’autonomous research machines’ which steer their own search refinements in response to our requests for new materials. In this way, solid freeforming used for sample preparation can give an ’arm’ to an intelligent machine so that it can conduct its own experimentation and learning; an idea that originated with Alan Turing in the late 1940s.     

Operation safety risk analysis method of hydropower project considering time dependent effect

In order to consider the time-dependent characteristic of risk factors of hydropower project,the method of stochastic process simulating structure resistance and load effect is adopted.On the basis of analyzing the structure characteristics and mode of operation,the operation safety risk rate assessment model of hydropower project is established on the comprehensive application of the improved analytic hierarchy process,the time-dependent reliability theory and the risk rate threshold.A scheme to demonstrate the time-dependent risk rate assessment method for an example of the earth-rock dam is particularly implemented by the proposed approach.The example shows that operation safety risk rate is closely related to both the service period and design standard;considering the effect of time-dependent,the risk rate increases with time and the intersection of them reflects the technical service life of structures.It could provide scientific basis for the operation safety and risk decision of the hydropower project by predicting the trend of risk rate via this model.     

一种对IP拥塞控制算法的改进方法

针对随机及早检测(RED)中存在的不稳定现象,提出了一种新颖的AQM算法LRC-RED,该模式具有快的响应速度、良好的鲁棒性以及高的链路利用率.LRC-RED是在RED,LossRadio,VRC,TCP/AQM等基础上提出的,它检查最近包的丢失率以及当前拥塞链路总速率,然后动态调整包的丢失概率以提高服务性能.为了验证算法的有效性,设计了一个有效的实验来检验.实验表明,该算法是可行的.     

RED-PD算法的仿真研究

随机早期检测(RED:Random Early Detection)是IETF推荐的一种基于路由器有效的主动队列管理算法,但是在某些情况下,一些数据量很大的数据流会大量占用带宽,从而导致了各流量之间带宽分配的不公平性,甚至产生拥塞崩溃。对M ahajan提出的一种基于RED分组丢弃历史的AQM(Active Queue M anagem ent)算法RED-PD(RED w ith Preferential D ropp ing)进行了深入研究。该算法通过对被检测出的高带宽流的数据包采用提前丢弃的策略,实现带宽分配的公平性。通过仿真发现了RED-PD算法的自适应性,表明此算法有更高的理论和实用价值。     

RED算法的改进及其性能分析

随机早期检测（RED）算法作为主动队列管理（AQM）中有效的实现算法，其性能研究近来已成为网络研究 的一个热点。介绍了原始RED算法，并在此基础上对该算法进行了改进，仿真结果和实验数据表明，改进算法能 显著地降低丢包率，提高网络的链路利用率。     

RED算法的改进及其性能分析

随机早期检测(RED)算法作为主动队列管理(AQM)中有效的实现算法,其性能研究近来已成为网络研究的一个热点。介绍了原始RED算法,并在此基础上对该算法进行了改进,仿真结果和实验数据表明,改进算法能显著地降低丢包率,提高网络的链路利用率。     

基于路由队列管理提高VOIP通话质量的方法研究

为解决当网络发生拥塞时,VOIP通话效果急剧降低的问题,提出了一种无需改造现有网络结构,低成本提升VOIP通话质量的方法. 通过随机早期检测算法的应用,可以有效地降低VOIP通话的延迟、延迟抖动、有效丢包率和连续丢包率,进而提高VOIP的通话质量;而且,即使在网络拥塞的极端情况下,也能保证一定的通话质量(QoS).      

一种高容量视频点播服务器的体系结构及其原型实现

为了支持大量VOD用户，提出了一种用于高容量视频点播服务器的单系统控制多数据泵(SSCWMDP)体系结构，能在Internet环境中大规模地提供视频点播服务．为了验证该体系结构的有效性，实现了基于该体系结构的视频点播服务器的原型．实验结果表明，采用SSCWMDP体系结构，该原型能够同时支持大量的客户．     

一种基于平均负载的阈值控制AQM算法

提出一种新的AQM算法——基于平均负载的阈值控制算法TLED,并研究在低延时状况下的性能情况。使用指数加权平均(EWMA)方法计算网络的平均负载强度,根据平均负载对数据包进行随机丢弃,通过一种阈值控制算法,将缓存区队列长度与负载状况有机结合起来,取得了吞吐量-队列延时的良好折中,并有效改善了原有的基于负载的控制算法对网络流量变化的自适应性。     

基于神经网络逆系统的鲁棒主动队列管理算法

通过在中间节点上使用主动队列管理策略来进行有效地拥塞控制,在保证较高吞吐量的基础上稳定地控制队列长度,从而实现了端到端的时延控制和保证QoS需求.在研究中,TCP的流量控制过程被视为二阶非线性时变系统,并通过可逆分析,证明该系统可逆,采用神经网络逆系统这种近年来发展起来的非线性鲁棒控制理论作为控制器的设计方法,设计出一种新的主动队列管理算法.仿真试验表明,这种算法的稳态和瞬态性能都优于与其具有相同实现复杂度的 RED和PI算法,并且在负载扰动和参数变化时具有很强的鲁棒性.神经网络逆系统方法应用于非线性的流量控制过程中有助于系统稳定性和鲁棒性.     

网络系统的不确定自适应滑模控制

针对TCP网络的拥塞控制问题,采用滑模控制理论提出了一种新的主动队列管理算法。考虑到网络系统不确定性上界很难获得,根据改进的李亚普诺夫函数,提出了一种简单的自适应律以适应系统的不确定的上界,并根据此自适应律设计了一个滑模控制器。所设计的控制器既保证了滑动模态的存在和系统的渐进稳定性,又较好的抑制了系统不确定带来的影响。仿真结果表明该方法的有效性。     

分布式反馈分组调度算法DF2Q

路由器要求提供高速转发性能及高服务质量(Qo S)。支持 Qo S的算法对缓冲队列进行分组调度。输出排队机制可获得要求的 Qo S,却缺少可扩展性 ;输入排队方式扩展性好 ,但缺少必要的 Qo S特性。采用输入输出结合排队方式 ,设计实现了可保证较好可扩展性和 Qo S的分组调度算法 DF2 Q。该调度算法引入反馈机制 ,很好避免内部拥塞和提高资源利用率。 DF2 Q提供和输出端公平排队近似的Qo S保证 ,保证路由器转发性能