Network and Parallel Computing

The study resulting in this paper applied a parallel algorithm based on a fourth-order compact scheme and suitable for parallel implementation of scientific/engineering systems. The particular system used for demonstration in the study was a time-dependendent system solved in parallel on a 2-head-node, 224-compute-node

Apple Xserve G5

multiprocessor. The use of the approximation scheme, which necessitated discretizing in both space and time with

h

x

space width and

h

t

time step, produced a linear tridiagonal,

almost-Toeplitz

system. The solution used

p

processors with

p

ranging from 3 to 63. The speedups,

s

p

, approached the limiting value of

p

only when

p

was small but yieldd poor computations errors which became progressively better as

p

increases. The parallel solution is very accurate having good speedups and accuracies but only when

p

is within reasonable range of values.

How to guarantee user’s QoS (Quality of Service) demands becomes increasingly iportant in service-oriented grid environment. Current research on grid resource advance reservation, a well-known and effective me-chanism to guarantee QoS, fails to adapt to dynamic variability of grid resource, and imprecise deny of user’s reservation request often happens. For this, new system architecture for advance reservation is proposed. SNAP (Service Negotiation and Acquisition Protocol) is extended to support this architecture. Then VRC (virtual resource container) is adopted to alleviate negative effect resulted from resource performance variability, and QDD (QoS deviation distance) based logical resource selection algorithm is addressed to decrease imprecise reject rate of reservation. At last, this new architecture is deployed to campus grid, and two illustrative experiments are conducted in campus grid too. Preliminary results show that it can alleviate negative influence of grid resource dynamic fluctuation and avoid imprecise reject of advance reservation request effectively.

Proxy Certificates (PCs) is one of key mechanisms in Grid Security Infrastructure (GSI). Users need PCs to access grid service. But there is no effective mechanism to manage the PCs in GSI. In order to apply GSI in Military Grid, a novel adaptive Proxy Certificates management scheme is brought forward based on the hierarchical one-way hash chains. The hierarchical one-way chain consists of two or more levels of chains, where values of a first-level chain act as roots of a set of second-level chains and each PC is protected by a hash value, so the PCs’ available time can be controlled adaptively and safely. The experimental results indicate that the scheme more adapt to the Military Grid environments.

A notable requirement of clusters is to maximize its processing performance. Lots of work in this area has been done to optimize the system performance by improving certain metric such as reliability, availability, security and so on. However, most of them assumes that the system is running without interruption and seldom considers the system’s intrinsic characteristics, such as failure rate, repair rate and lifetime. In this paper, we study how to achieve high availability based on residual lifetime analysis for the repairable heterogeneous clusters with makespan constraints. First, we provide an availability model based on addressing the cluster’s residual lifetime model. Second, we give an objective function about the model and develop a heuristic scheduling algorithm to maximize the availability the makespan constraint. At last, we demonstrate these advantages through the extensive simulated experiments.

Due to high latency of the Internet, it becomes a challenge to access large and widely distributed data quickly and efficiently on data grids. Replication is a process to address this issue, by storing data in different locations to reduce access latency and improve data locality. However, because of the limited storage capacity, a good replica replacement algorithm is needed to improve the efficiency of the access to the replicas. In this paper, a VO-based two-stage replica replacement algorithm is proposed, which provides a good solution to deal with the relations between value and cost. In a VO, replica value is determined according to popularity to make sure which replica will be replaced, and the bandwidth is predicted to make replacement cost as low as possible. Experiments show that compared with traditional replacement algorithms our new replica replacement algorithm shows better performance and efficiency of the data access on Data Grids.

One of the biggest challenges in building grid schedulers is how to deal with the uncertainty in what future computational resources will be available. Current techniques for Grid scheduling rarely account for resources whose performance, reliability, and cost vary with time simultaneously. In this paper we address the problem of delivering a deadline based scheduling in a dynamic and uncertain environment represented by dynamic Bayesian network based stochastic resource model. The genetic algorithm is used to find the optimal and robust solutions so that the highest probability of satisfying the user’s QoS objectives at a specified deadline can be achieved. It is shown via a simulation that the new methodology will not only achieving a relatively high probability of scheduling workflow with multiple goals successfully, but also be resilient to environment changes.

In this paper, we propose a dynamic file transfer scheme with co-allocation architecture, called Dynamic Adjustment Strategy, a dynamic file transfer scheme with co-allocation architecture that reduce the file transfer times and improves the performance in Data Grid environments. Our approach reduces the idle time faster servers spend waiting for the slowest server, and decreases file transfer completion time. We also present a new toolkit, called Cyber-Transformer, with a friendly graphical user interface interface running on the client side that makes it easy for inexperienced users to manage replicas and download the files in Data Grid environments. We also provide an effective scheme for reducing the cost of reassembling data blocks.

A distributed map data parallel dissemination system, which can be used to distribute spatial map data to remote users through different networks, is presented. The paper first defines spatial map data, associated metadata and network resource nodes with graphic formalization definition. Then, the relations between map data and network resources are established. Based on formalized definition to the whole system, a map data dissemination framework and a series of essential distributed CORBA services, conditions input methods based on three querying conditions are presented. We also explore network map copyright validating in dissemination course, and present an improved adaptive watermarking algorithm for vector digital maps whose data is the most difficult and necessary to watermark. Different from other methods, we compare the watermarked map with the original watermarked map, not with the primitive map, before extraction. The results show that compared with Direct Copy, our dissemination system can delivery spatial map data effectively. The transfer performance of CORBA services mode is almost equal to Direct Copy. The system can approach the biggest transfer rate more quickly but dependents on data amount weakly. And, the improved watermarking algorithm has better robustness.

Email overload is a recent problem that there is increasingly difficulty people have faced to process the large number of emails received daily. Currently this problem becomes more and more serious and it has already affected the normal usage of email as a knowledge management tool. It has been recognized that categorizing emails into meaningful groups can greatly save cognitive load to process emails and thus this is an effective way to manage email overload problem. However, most current approaches still require significant human input when categorizing emails. In this paper we develop an automatic email clustering system, underpinned by a new nonparametric text clustering algorithm. This system does not require any predefined input parameters and can automatically generate meaningful email clusters. Experiments show our new algorithm outperforms existing text clustering algorithms with higher efficiency in terms of computational time and clustering quality measured by different gauges.

Multi-log

2

N

networks architecture is attractive for constructing optical switches, and the related routing algorithms are critical for the operation and efficiency of such switches. Although several routing algorithms have been proposed for multi-log

2

N

networks, a full performance comparison among them is not available by now. Thus, this paper is committed to such a comparison in terms of blocking probability, time complexity, hardware cost and load balancing capability. Notice that the load balance is important for reducing the peak power requirement of a switch, so we also propose in this paper two new routing algorithms for optical multi-log

2

N

networks to achieve a better load balance.

Vertically stacked optical banyan (VSOB) is an attractive architect-ture for constructing banyan-based optical switches. Blocking analysis is an effective approach to studying network performance and finding a graceful compromise among hardware cost, blocking probability and crosstalk tolerance; however, little has been done on analyzing the blocking behavior of VSOB networks under crosstalk constraint. In this paper, we study the overall blocking behavior of a VSOB network under various degree of crosstalk, where an upper bound on the blocking probability of the network is developed. The upper bound depicts accurately the overall performance behavior of a VSOB network as verified by extensive simulation results and it agrees with the strictly nonblocking condition of the network. The derived upper bound is significant because it reveals the inherent relationship between blocking probability and network hardware cost, by which a desirable tradeoff can be made between them under various degree of crosstalk constraint. Also, the upper bound shows how crosstalk adds a new dimension to the theory of switching systems. In particular, our bound provides network developers an effective tool to estimate the maximum blocking probability of a VSOB network in which different routing algorithms can be applied with a guaranteed performance in terms of blocking probability and hardware cost. An important conclusion drawn from our work is that the hardware cost of VSOB networks can be reduced dramatically without introducing significantly high blocking probability considering the crosstalk.

In this paper, we propose a new structured P2P overlay network, named SW-Uinta, where employs a non-deterministic caching strategy that allows for polylogarithmic search time while having only a constant cache size. Compared with deterministic caching strategies proposed by previous P2P systems, the non-deterministic caching strategy can reduce communication overhead for maintaining the routing cache table. Cache entries in the peer can be updated by subsequent queries rather than only by running stabilization periodically. A novel cache replacement scheme is used to improve lookup performance. We compare the performance of our system with that of other structured P2P networks such as Chord and Uinta. It shows that the SW-Uinta protocol can achieve improved object lookup performance and reduce maintenance cost compared with some other protocols.

In order for a vehicle to follow a predetermined trajectory accurately, its position must be estimated accurately and reliably. In this paper, we propose new lateral control methods for unmanned vehicles and a positioning system using ultrasonic waves. The positioning problem is considered as an important issue of control problem for unmanned navigation of a vehicle. Dead Reckoning is widely used for positioning of vehicle. However this method has problems because it accumulates estimation errors. We propose a new method to increase the accuracy of position estimation using the Ultrasonic Satellite system. It is shown that we will be able to estimate the position of vehicle precisely, in which errors are not accumulated. We also propose new lateral control methods including a new path planning method and a heading angle modulator. The experimental results show that the proposed methods enable accurate vehicle trajectory tracking under various environmental factors.

Slowdown factors determine the extent of slowdown a computing system can experience based on functional and performance requirements. Dynamic Voltage Scaling (DVS), which adjusts the clock speed and supply voltage dynamically, is an effective technique in reducing the energy consumption of embedded real-time systems. We address the problem of computing static and dynamic slowdown factors in the FPPT algorithm. In this paper, Sufficient constraints have been identified for the feasibility of the task set using slowdown factors. We formulate this problem of computing the static slowdown factors for tasks as an nonlinear optimization problem to minimize the total energy consumption of the system. Our simulation experiments show on an average 17%~53% energy gains over FPPT scheduling policy.

This paper proposes a new method to find an absolute position by using ultrasonic sensors. In order to evaluate the performance of U-SAT (Ultrasonic Satellite system), the autonomous navigation performance of a mobile robot is tested. Experiments were performed in both cases that the mobile robot moves to the target point using relative positioning method in conjunction with U-SAT, which is, absolute positioning methods. The performance of U-SAT is evaluated accordingly with the results of the experiments. As a result, U-SAT could be effectively used as a pseudolites or pseudo-satellites to help a mobile robot navigate intelligently and autonomously in an indoor area.

Service sharing and discovery play a relevant role in mobile ad hoc environments. Upon joining a self-organizing network, mobile nodes should be able to explore the environment to learn about, locate, and share the available services. In this paper, we propose a distributed and scalable service discovery and sharing framework for ad hoc networks. The proposed framework defines three types of nodes: service directories, service providers and requesting nodes. Service directory nodes act as mediators for lookup requests from requesting nodes. Joining service provider nodes register their services with the nearest service directory. A requesting node discovers the available services by submitting requests to its nearest service directory which determines the node providing the requesting service. The performance of the proposed model is evaluated and compared to the broadcast-based model that has been extensively studied in the literature.

Reading the contents of Web page with a small-screen device, such as a PDA or cell-phone, is still far from being a pleasant experience. Owing to the device limitations, current mobile browsers cannot handle all HTML tags, such as tables, for instance. Thus, most mobile browsers provide a linearized version of the source HTML page, leading to a large amount of scrolling, not to mention the difficulty in finding the desired content. The main contribution of this work is to propose an architecture for adapting web page on small-screen devices. Among the features that our architecture offers, we can cite on-the-fly Web page adaptation and customization according to the user and device characteristics; text summarization; page blocks identification and content mapping to easy the task of locating user interests.

The geographic routing technology is good for the self-organizing and large-scale Wireless Sensor Networks(WSNs), and the energy-limited sensor nodes require the energy used for routing to be minimum. In this paper a new energy-efficient geographic routing algorithm is proposed and it is distributed and based on the geographic routing, the topology characteristics of the network and the wireless communication energy model. The algorithm is based on the planarized graph (GG) of the network, it deals with the routing void by saving more face neighbors in every node, and selects the most energy-efficient nexthop node by using the energy-saving technologies including the Minimum Energy One-hop Neighbor Path Selection and the Optimal Face-neighbor Selection. The theoretical analysis and simulations show that the algorithm is feasible and more energy-efficient than many existed geographic routing algorithms. In the end , the means of face information maintenance are proposed.

Currently, people can collect information and share resources on the Internet through WiFi. The Voice over Internet Protocol (VoIP) supported by wireless technology has made Internet access more versatile and flexible than before. However, the more sessions established for communication, the longer the transmission delay and the higher the packet dropping rate. In this paper, an improved bandwidth utilization approach developed on the IEEE 802.11e standard is proposed. Experimental results show that this approach significantly improves a system’s throughput as compared with 802.11b and other schemes.

Added delay strategy can be used to solve the broadcast storm problem of ordinary broadcast mechanism (OBM) and maximize network life-time. Available added delay strategies take into account the distance and/or the residual energy. In this paper, we propose a new added delay strategy—Maximum Life-time Localized Broadcast (ML

2

B). As the node’s number of neighbors that have not received the broadcast message (we call it coverage degree) can better describe the coverage rate, ML

2

B takes the coverage degree rather than the distance into account. ML

2

B also takes the residual energy into account as other strategies do. ML

2

B only need one-hop neighbor information to find the coverage degree, so ML

2

B is a distributed protocol and the overhead is small. Simulation results show that, as compared with OBM, ML

2

B can save at least 50% rebroadcast, its maximum end-to-end delay is lower, its reachability is the same, and its network life-time is two times longer.

In this paper, we propose a modulation multiplexing distributed space-time coding for two-user cooperative diversity in wireless network, in which the information of one user lies in the in-phase axis (

I

axis), while that of the other user lies in quadrature-phase axis (

Q

axis). Since two users share both time and frequency resource in cooperation sub-frame, the proposed scheme is more bandwidth efficient. We characterize performance of the proposed scheme in symmetric inter-user case. Simulation results show that the proposed scheme outperforms the amplify-to-forward cooperative system. Compared to the distributed space-time coding with time-division multiple accesses cooperative system, the proposed scheme achieves the same diversity gain but loses some coding gain. However, when the selection relay protocol is adopted for the proposed scheme, coding gain is also achieved.

Widest Disjoint Paths (

WDP

) algorithm is a promising multipath routing algorithm aimed at selecting good paths for routing a flow between a source-destination pair, where their bottleneck links are mutually disjoint. Nevertheless, the complexity of

WDP

algorithm is relatively high due to the fact that the good path selection process considers all available paths. To reduce its complexity, This paper proposes a modified

WDP

algorithm, which uses only a subset of available paths based on shortest widest paths, thereby limiting the number of candidate paths considered. As a result, the number of iterations in the good path selection process is significantly reduced. Performance analysis shows the modified scheme is more efficient than the original algorithm in a large network. Simulation results demonstrate that, in comparison with the original

WDP

algorithm, the modified

WDP

algorithm leads to lower latency and faster packets transferring process as the number of available paths increases.

In a multiprocessor network, sending a packet typically refers to start-up time and transmission time. To optimize these two times, as opposed to earlier solutions, a spanning tree and multiple spanning trees are constructed to solve four types of broadcasting problems in an (n, k)-star graph: one-to-all or all-to-all broadcasting with either one-port or all-port communication model, respectively. Since the proposed spanning tree has an optimal height, both one-to-all and all-to-all broadcasting algorithms achieve nearly optimal start-up time and transmission time under all-port model and one-port model, and optimal transmission time under one-port model. By using multiple spanning trees, both one-to-all and all-to-all algorithms achieve nearly optimal transmission time under all-port model and one-port model.

To fulfill the application of wireless network in small area, this paper proposed a protocol for decentralized all-connected wireless network. This protocol is based on DS-CDMA with the ability of multi-user detection (MUD). And it has the characteristics of transmitting data in parallel, high network throughput and low communication error rate. In this paper, the process of the protocol is introduced, and an improved quasi-synchronous decorrelation multi-user detection algorithm is deduced in detail. And then, the protocol is simulated and the results are analyzed.

In this paper, by introducing knowledge of fuzzy mathematics, probability theory and gaming theory, a QoS unicast routing scheme with ABC supported is proposed based on small-world optimization algorithm. Under inaccurate network status information and imprecise user QoS requirement, the proposed scheme uses the range to describe the user QoS requirement and the edge parameter, introduces the user satisfaction degree function, the edge evaluation function and the path evaluation function, trying to find a QoS unicast path with Pareto optimum under Nash equilibrium on both the network provider utility and the user utility achieved or approached. Simulation results have shown that it is both feasible and effective with better performance.

An important issue in deploying a wireless sensor network (WSN) is to provide target coverage with high energy efficiency and fault-tolerance. In this paper, we study the problem of constructing energy-efficient and fault-tolerant target coverage with the minimal number of active nodes which form an

m

-coverage for targets and a

k

-connected communication subgraph. We propose two heuristic algorithms for

m

-coverage problem, and get the performance ratio of one heuristic. Then two heuristic algorithms are further proposed to solve the

k

-connected

m

-coverage problem. The simulation results demonstrate the desired efficiency of the proposed algorithms.

Based on the concept of random contention and collision resolution, a QoS-based multiple access (QMA) protocol for ad hoc networks is proposed to support multimedia service and multi-hop architecture. In this paper, the traffic is divided into two groups with different QoS requirements, namely real-time traffic and non real-time traffic. According to the protocol, nodes with real-time traffic have higher priority to access channel than those with non real-time traffic by broadcasting forecast bursts (FB) in earlier contention slots. Meanwhile, real-time traffic is scheduled according to its delay and the earliest deadline first (EDF) principle. Through simulations, it is shown that the QMA protocol outperforms the Carrier Sensing Multiple Access with Collision Avoidance (CSMA/CA) protocol in terms of throughput, message discard rate and average packet delay, and the QMA protocol can provide differentiated QoS guarantees for traffic in multi-hop networks.

The paper introduces a new approach based on dual residue system to compute Montgomery multiplication. The novelty of this proposal is that we import an extra Montgomery residue system with new transformation constant beside the normal one. In this way, one of the multiplicand can be divided into two parts and both higher and lower parts are calculated in parallel to speed up computation. Then two implementations in hardware are proposed for the algorithm. In parallel architecture, the proposed algorithm can perform nearly twice speedup compared to normal Montgomery method. And in pipeline architecture, the computation speed can be even faster. Besides speeding up calculation the extra merit of our proposal is that the multiplier can partial replace Montgomery multiplier used nowadays without any changes on top architecture.

This paper analyzes the conflict between performance enhancing technology and IPSec in satellite IP networks, and proposes a solution called multilayer IP security with changeable zone (CZML-IPSec). It enables licensed intermediate nodes not only access TCP header, but also object links of upper layer in the form of HTML by converting static zone mapping to changeable dynamic mapping and building up composite security association correspondingly. A prototype is implemented to demonstrate the practical feasibility of CZML-IPSec. Measurements and performance analysis indicate that CZML-IPSec does not add unacceptable bandwidth overheads and delay, and it does not increase substantially processing hardware requirements. CZML-IPSec can help satellite IP networks provide both end-to-end security and performance enhancement.

This paper proposes a novel scheme of group key management based on Jacobian Elliptic Chebyshev Rational Map, named Jacobian Group Key Management(JGKM). The scheme is more efficient than other group key managements since fewer re-keying messages are sent when group membership changes. Besides, it provides both forward and backward secrecy. Therefore, this proposal is helpful to deploy secure multicast over some networks with high latency or limited bandwidth such as wireless network. Furthermore, it fits both small-scale and large-scale groups.

A scheme that defending against distributed denial of service (DDoS) attacks adopts the mechanism of Distribution-based Secure Overlay Nodes (DSON) to a large-scale ISP (Internet Service Provider) network is presented. The scheme uses local BPG announcement to divert traffic to the overlay network when experiencing high load, then filtering algorithm based on the technology of signal processing is applied to the diverted traffic. This algorithm detects and filters out DDoS attacks in frequency domain to allow targets to provide good service to legitimate traffic, with fast reaction and high energy ratio of legitimate to attacks traffic. DSON is implemented and installed on the monitor points of large-scale ISP network associated with the corresponding routers, edge router, border router, and core router, with no requirement for the modifying to network architecture, infrastructure, and protocol.

With the Canetti-Krawczyk (CK) model, we analyze the authentication module WAIs in the Chinese WLAN national security standard WAPI and its implementation plan respectively. The security weaknesses of WAI in the original WAPI are presented; then WAI in the implementation plan is proved secure in the CK model; at last we point out how the implementation plan overcomes the security weaknesses in the original WAPI.

Large IP backbone networks today are mostly deployed directly over sequences of point-to-point DWDM systems or chains of newer ROADM-based ultra long haul systems, interconnected by OEO regenerators. The next generation core optical network is moving toward an all-optical network architecture that is based on multi-degree ROADMs to reduce OEO regeneration cost as well as enabling automatic reconfigurability and dynamic restoration. In this paper, we study the restoration design in this new IP over reconfigurable all-optical network architecture to satisfy the resilience requirements for both IP and wavelength services. We propose two novel restoration schemes: 2-Phase Fast Reroute mechanism with optimized Traffic Engineering algorithm for restoring IP services and shared mesh restoration with standbys for restoring wavelength services. They both meet the requirement of sub-second restoration time and also maximize sharing among different failures with the objective of minimizing either overall capacity or overall cost. To further reduce the required restoration capacity in both IP layer and optical layer and address failures in both layers efficiently, we also propose an integrated IP-over-optical layer restoration strategy that enables sharing of restoration capacity among non-simultaneous failures across both IP and optical layers. Simulation results demonstrate significant improvements using our proposed schemes comparing with existing ones.

In the fast-growing internet applications, email becomes more and more important in communication. SMTP attacks and spam have become one of the most serious problems. Particularly, the SMTP attacks and spam varies on email, for example spoofing address, illegal characters, sending in bulk, too many SMTP commands and so on. A single security technique is not enough to protect the system from these attacks and spam. In this paper, we propose a SMTP Intrusion Prevention System (SIPS) which bases on the concept of Stateful Protocol Anomaly Detection and Flow-based Inspection. SIPS is implemented by a finite state machine to inspect all coming email flows. It is according to the media type of email flow and their characteristics. On the test of a real email environment, our approach can prevent attacks on SMTP attack (mail bomb) average about 95.4% and spam average about 91.1%.

Network security is very important for Internet-connected hosts because of the widespread of worms, viruses, DoS attacks, etc. As a result, a network intrusion detection system (NIDS) is typically needed to detect network attacks by packet inspection. For an NIDS system, string matching is the computation-intensive task and hence the performance bottleneck, since every byte of the payload of packets must be checked against numerous predefined signature strings, which may occur arbitrarily in the payload. In this paper, we present the design and evaluation of parallel string matching algorithms targeting hardware implementation on FPGAs and software implementation on multi-core processors. Experimental results show that, on a multi-processor system, the multi-threaded implementation of the proposed parallel string matching algorithm can reduce string matching time by more than 40%.

The current storage models for Object-Oriented DataBase (OODB), which organize data as objects according to the Object-Oriented Data Model (OODM), are mainly established on the block storage devices. In this way, the storage manager does not have detailed knowledge of the characteristics of the underlying storage devices, and the storage subsystem does not have the semantic knowledge of the data stored in the block storage devices, so it is very difficult to implement some workload-dependent tasks such as data layout and caching. Furthermore, the storage subsystem of OODB has to organize the objects in the pages, which is not well-suited to the objects storage. In this paper, we present an Object-Based Storage Model (OBSM) for OODB by using the recently-standardized Object-based Storage Device (OSD) interface. OSD offloads the storage management into the storage device itself and provides an object interface to data, which brings more intelligence into the storage subsystem. In the first glance at using OBSD in OODB, we explore the methods to map OODM to OBSM including direct mapping, mapping clustering to collection and declustering a large object to a series of sub-objects, and analyze the benefits to the storage subsystem of OODB by using OBSM such as providing storage functionalities offloading, providing objects sharing pool, providing integrative object persistence.

In this paper a high performance storage system for RDF documents is introduced. The system employs optimized index structures for RDF data and efficient RDF query evaluation. The index scheme consists of 3 types of indices. Triple index manages basic RDF triples by dividing original RDF graph into several sub-graphs. Path index manages frequent RDF path patterns for long path query performance enhancement. Context index is optional for context oriented RDF data and temporal RDF data. In this paper, we describe the organization of index structures, show the process of evaluating queries based on the index structures, and provide a performance comparison with exist RDF databases through several benchmark experiments.

Almost all cryptographic file sharing systems need re-encryption when the sharing was revoked. These systems differ from each other only in the timing of re-encryption. As re-encryption is an expensive operation, it is significant to avoid re-encryption. The purpose of this paper is to advise a direction to avoid re-encryption and facilitate file sharing in cryptographic file sharing system. A Black-box model is set up to achieve this objective. In the model, FPGA or ASIC chips are used to act as the black-box as they have been extensively researched and applied in cryptography. Some applications of FPGA and ASIC in cryptography are detailed in this paper. Their feasibility to be functioned as the black-box is discussed. Also a software implementation on FPGA is attached with tested and effective performance.

By run-time characteristic analysis of parallel workloads, we found that a majority of shared data accesses of parallel workload has temporal locality. Based on this characteristic, we present a sharing relation cache (SRC for short) based CMP architecture, saving recently used sharing relations to provide destination set information for following cache-to-cache miss requests. Token-SRC protocol integrates SRC into token protocol,reducing network traffic of token protocol.Simulations using SPLASH-2 benchmarks show that, a 16-core CMP system with token-SRC achieved average 15% network traffic reduction of that with token protocol.

Network processors are specialized integrated circuits used to process packets in such network equipment as core routers, edge routers, and access routers. As predicted by Gilder’s law, Internet traffic has doubled each year since 1997 and this trend is showing no signs of abating. Since all emerging network applications which require deep packet classification and security-related processing should be run at line rates and since network speed and network applications complexity continue increasing, future network processors should simultaneously meet two requirements: high performance and high programmability. Single processor performance will not be sufficient to support the requirements which will be imposed on future network processors. In this paper, we consider the CMP model as the baseline architecture of future network processors. We investigate the architectural implications of cache coherence protocols with network workloads on CMPs. Our results show that the token protocol which uses the tokens to control read/write permission of shared data blocks shows better performance than the directory protocol by a factor of 13.4%.

This paper proposes a semi-completely-connected bus, called

SKB

, to alleviate the long-wire and pin-neck problems against on-chip systems through a small diameter and dynamic clustering. Dynamic clustering allows to reduce the traffic to the per-cluster units such as the global interconnect interface, as compared with the static clustering fixed in hardware. We derive a 2

n

-node

semi-complete (SK) graph

from a simple node-partitioning. An SKB is produced from the SK graph when we replace the links incident to a node by a single bus for the node. The diameter of SKB equals 1 (bus step), though the bus length is rather long,

O

$(\sqrt{2^{n}})$

. Simulation results show that relative to the hypercube with the link delay of 1 clock, the SKB’s bandwidth is about 0.97 and 0.14 assuming the bus delay of 1 and 8 clocks, respectively, that increases to about 4.57 and 0.71 with the dynamic clustering.

Java is widely applied into embedded devices. Java programs are compiled into Java bytecodes, which are executed into the Java virtual machine. The Java virtual machine is a stack machine and instruction folding is a technique to reduce the redundant stack operations. In this paper, a simple instruction folding algorithm is proposed for a Java processor named jHISC, where bytecodes are classified into five categories and the operation results of incomplete folding groups are hold for further folding. In the benchmark JVM98, with respect to all stack operations, the percentage of the eliminated P and C type instructions varies from 87% to 98% and the average is about 93%. The reduced instructions are between 37% and 50% of all operations and the average is 44%.

Network distance is an important parameter in optimizing performance of network applications. Although there are a number of network distance prediction mechanisms, they all take no consideration of Internet structure, which has great influence on Internet distance characteristics. By analyzing the hierarchical structure feature of Internet, a hierarchical network distance predication mechanism called HNDP is proposed. HNDP divides Internet into many independent prediction regions, and predicts distance information between network nodes by accumulating distances in different predication regions, which can avoid the problem that short distance and long one cannot be accurately predicated simultaneously. To optimize the influence of landmark selection on HNDP prediction accuracy, a shortest distance cover landmark selection model is proposed, and then a tabu search algorithm called TS_Landmark is given to solve this model in HNDP. Finally, the simulation results under ns-2 show that TS_Landmark can select landmarks effectively, and HNDP provides more accurate results than traditional single layer ones.

We analyze the MAC performance of the IEEE 802.15.4 LR-WPAN non-beacon mode with the piggyback method in non-saturated condition. Our approach is to model a stochastic behavior of one device as a discrete time Markov chain. We propose an analytical model describing the download behavior of a device using piggyback method. We obtain the performance measures such as throughput, packet delay, energy consumption and packet loss probability of a device. Numerical results and simulation results show that the piggyback method which removes a backoff procedure in the backoff method can reduce the delay, loss probability and energy consumption compared with backoff method. Our results can be used to find the optimal number of devices with some constraints on packet delay and packet loss probability.

Traffic sampling technology has been widely deployed in front of many high-speed network applications to alleviate the great pressure on packet capturing.Increasingly passive traffic measurement employs sampling at the packet level. Packet sampling has become an attractive and scalable means to measure flow data on high-speed links. However, knowing the number and length of the original flows is necessary for some applications. This paper provides a novel algorithm, Least Square Method(LSM), that uses flow statistics formed from sampled packet stream to infer the absolute frequencies of lengths of flows in the unsampled stream. The theoretical analysis shows that the computational complexity of this method is well under control, and the experiment results demonstrate the inferred distributions are as accurate as EM algorithm.

Distributed applications running on clusters may be composed of several components with very different performance requirements. The FlowVR middleware allows the developer to deploy such applications and to define communication and synchronization schemes between components without modifying the code. While it eases the creation of mappings, FlowVR does not come with a performance model. Consequently the optimization of mappings is left to the developer’s skills. But this task becomes difficult as the number of components and cluster nodes grow and even more complex if the cluster is composed of heterogeneous nodes and networks. In this paper we propose an approach to predict performance of FlowVR distributed applications given a mapping and a cluster. We also give some advice to the developer to create efficient mappings and to avoid configurations which may lead to unexpected performance. Since the FlowVR model is very close to underlying models of lots of distributed codes, our approach can be useful for all designers of such applications.

The switching technique determines how messages are propagated from source to destination, and has a great impact on network performance. Traditional flow control mechanisms such as Wormhole Switching (WS) realize very good performance, but prone to deadlock in the vicinity of faults. While techniques such as adaptive routing can alleviate the problem, it cannot by itself solve the problem. This has motivated the development of different switching techniques. The Scouting Switching (SS) has been suggested as an efficient switching method for reconciling the conflicting demands of communication performance and fault-tolerance in computer networks. In this paper, we present a novel mathematical model to predict communication delay of SS coupled with virtual channels and fully adaptive routing in 2-D torus networks. We have carried out extensive simulation experiments, the results of which are used to validate the proposed analytical model.

The multiply clock domain (MCD) technique is a novel technique to compromising between synchronous systems and asynchronous systems to reduce the power. However, most present studies of MCD are based on superscalar architectures. In this paper, MCDE, a MCD technique based on explicitly parallel instruction computing (EPIC) architecture is designed and implemented to reduce the power of clock distribution network. In addition, a series of experiments have been done to evaluate it. The result of the experiments show that, using a MCDE clock network microarchitecture with a fine-grained adaptive dynamic adjustment algorithm, can effectively decrease the microprocessor power by 40%, compared with the original EPIC clock network microarchitecture.

The continual improvement of service process is one of the most important factors for improving service process management level. But it is very difficult to identify the real causes of service fault executing service process improvement. Aiming at the problem, this paper proposed a service fault recognition & improvement method based on association rules mining. The method provides a new approach for enterprises to improve effectively service process fault.

In order to classify data with noises or outliers, Fuzzy support vector machine (FSVM) improve the generalization power of traditional SVM by assigning a fuzzy membership to each input data point. In this paper, an improved FSVM based on vague sets is proposed by assigning a truth-membership and a false-membership to each data point. And we reformulate the improved FSVM so that different input points can make different contributions to decision hyperplane. The effectiveness of the improved FSVM is verified in credit rating; the experiment results show that our method is promising.

Parallelism is one of the main sources for performance improvement in modern computing environment, but the efficient exploitation of the available parallelism depends on a number of parameters. Determining the optimum number of threads for a given data parallel loop, for example, is a difficult problem and dependent on the specific parallel platform. This paper presents a learning-based approach to parallel workload allocation in a cost-aware manner. This approach uses static program features to classify programs, before deciding the best workload allocation scheme based on its prior experience with similar programs. Experimental results on 12 Java benchmarks (76 test cases with different workloads in total) show that it can efficiently allocate the parallel workload among Java threads and achieve an efficiency of 86% on average.

This paper focuses on parallel interactive applications ranging from scientific visualization, to virtual reality or computational steering. Interactivity makes them particular on three main aspects: they are endlessly iterative, use advanced I/O devices, and must perform under strong performance constraints (latency, refresh rate). In this paper, we propose an application description language based on a data flow and hierarchical component model to cope with the complexity of parallel interactive applications. It enables us to define highly generic components, enforcing the application maintainability and portability. An implementation on top of the FlowVR middleware is presented.

Reconfigurable computing has become a hot topic in research due to its high-performance and flexibility. In this paper we present a simulator called JRSim for mesh-based reconfigurable architectures. The purpose of this simulator is to provide a platform to evaluate new architectures, and to assist in analysis of algorithms as well as the visualization of their behavior. JRSim is a platform-independent tool which is implemented by Java. It supports flexible bus structure, user-defined function unit and dynamic reconfiguration. Case studies show that JRSim can simulate the behavior of mesh-based reconfigurable systems correctly and efficiently. This simulator can be used to evaluate reconfigurable system design, or demonstrate the ability of reconfigurable system in an educational environment.