片上网络中面向报文有序传输的自适应路由算法研究

与确定性路由算法相比,自适应路由算法可以提高片上网络的通信性能,但是报文可能会无序到达。在目的节点对报文排序将会导致严重的面积和计算开销,甚至可能会抵消采用自适应路由算法带来的性能增益。为此,本文首先提出一种部分自适应路由算法,以满足报文的有序到达。然后,描述了对本文算法提供支持的路由器硬件结构。最后,在二维片上网络下对本文算法及确定性和自适应路由算法进行了性能评估和比较。与XY算法相比,本文算法显著降低了报文延时,提升了饱和点。同时讨论了对路由器面积和功耗影响。虽然路由器的功耗有所上升,但是由于报文交付性能提升,因此每个flit的能耗增长可忽略不计。     

高性能互连网络中的MSFS集成调度算法

提出最大服务优先（MSFS）的组播调度算法,从减小组播报文在单个路由器上传输延迟的角度出发,根据接收到报文副本的目标端口数和报文在缓存队列头位置的等待时间确定组播报文的优先级。MSFS实现简单,当单播和组播并存时,能够与FCFS单播调度算法进行“无缝”集成。模拟结果表明,MSFS能够取得高效、稳定的通信性能。     

PowerTrust: A Robust and Scalable Reputation System for Trusted Peer-to-Peer Computing

We propose a new fair scheduling technique, called OCGRR (output controlled grant-based round robin), for the support of DiffServ traffic in a core router. We define a stream to be the same-class packets from a given immediate upstream router destined to an output port of the core router. At each output port, streams may be isolated in separate buffers before being scheduled in a frame. The sequence of traffic transmission in a frame starts from higher-priority traffic and goes down to lower-priority traffic. A frame may have a number of small rounds for each class. Each stream within a class can transmit a number of packets in the frame based on its available grant, but only one packet per small round, thus reducing the intertransmission time from the same stream and achieving a smaller jitter and startup latency. The grant can be adjusted in a way to prevent the starvation of lower priority classes. We also verify and demonstrate the good performance of our scheduler by simulation and comparison with other algorithms in terms of queuing delay, jitter, and start-up latency     

OCGRR: A New Scheduling Algorithm for Differentiated Services Networks

We propose a new fair scheduling technique, called OCGRR (Output Controlled Grant-based Round Robin), for the support of DiffServ traffic in a core router. We define a stream to be the same-class packets from a given immediate upstream router destined to an output port of the core router. At each output port, streams may be isolated in separate buffers before being scheduled in a frame. The sequence of traffic transmission in a frame starts from higher-priority traffic and goes down to lower-priority traffic. A frame may have a number of small rounds for each class. Each stream within a class can transmit a number of packets in the frame based on its available grant, but only one packet per small round, thus reducing the intertransmission time from the same stream and achieving a smaller jitter and startup latency. The grant can be adjusted in a way to prevent the starvation of lower priority classes. We also verify and demonstrate the good performance of our scheduler by simulation and comparison with other algorithms in terms of queuing delay, jitter, and start-up latency.     

NLM: network-based layered multicast for traffic control of heterogeneous network

Layered multicast has been considered to be an efficient approach for coping with the network heterogeneity. Receiver-driven Layered Multicast (RLM) and its variants suggest that every receiver adjusts the reception level based on the packet loss rate. In this paper we propose a scheme called network-based layered multicast (NLM) scheme, which enhances the packet scoping mechanism employed in current multicast backbone (MBone). It makes use of the TTL bits and TOS bits in IP header, and packets are filtered in the routers using both the measured and predicted traffic condition. This allows the network quickly and efficiently adapts to the traffic condition. We evaluate and compare the scheme with the receiver-driven schemes using Network Simulator for a 33-node network topology with realistic bandwidths. The simulation reveals that the proposed NLM scheme substantially improves the stability and fairness of the RLM scheme for multicast sessions, while it displays comparable performance to the RLM with Priority (RLMP) scheme. For mixed type traffic of multicast and TCP sessions, it outperforms both the schemes.     

Distributed route repair for increasing reliability and reducing control overhead for multicasting in wireless MANET

The Multicast Ad hoc On-Demand Distance Vector (MAODV) routing protocol is proposed for achieving multicast in a Mobile Ad hoc Network (MANET) while reducing bandwidth waste and energy power consumption. In MANET, packets transmission through a multicast tree may always have unreliable links caused by node mobility or lack of energy, and thus significantly degrades the performance. MAODV uses a broadcast-type local repair mechanism to find an alternative route to the multicast tree when some breaks happen on the tree. Although the local repair mechanism provides a specified time-to-live (TTL) to limit the repair range and the hop-count to the group leader, a large number of broadcast-type Route Request (RREQ) messages extensively yields control overhead and requires a large amount of power consumption to send control messages. Thus, this paper proposes a unicast-type multihop local repair protocol for multicast MANETs to recover lost links efficiently while achieving several advantages: increasing network reliability, increasing packet delivery rate, minimizing the number of control messages and reducing repair delay. Moreover, the optimal number of hops used in the multihop neighbor table is analyzed mathematically. Numerical results indicate that the proposed approach outperforms other repair approaches in terms of successful repair rate, control message overhead and packet delivery rate.     

A NoC-based simulator for design and evaluation of deep neural networks

The astonishing development in the field of artificial neural networks (ANN) has brought significant advancement in many application domains, such as pattern recognition, image classification, and computer vision. ANN imitates neuron behaviors and makes a decision or prediction by learning patterns and features from the given data set. To reach higher accuracies, neural networks are getting deeper, and consequently, the computation and storage demands on hardware platforms are steadily increasing. In addition, the massive data communication among neurons makes the interconnection more complex and challenging. To overcome these challenges, ASIC-based DNN accelerators are being designed which usually incorporate customized processing elements, fixed interconnection, and large off-chip memory storage. As a result, DNN computation involves large memory accesses due to frequent load/off-loading data, which significantly increases the energy consumption and latency. Also, the rigid architecture and interconnection among processing elements limit the efficiency of the platform to specific applications. In recent years, Network-on-Chip-based (NoC-based) DNN becomes an emerging design paradigm because the NoC interconnection can help to reduce the off-chip memory accesses while offers better scalability and flexibility. To evaluate the NoC-based DNN in the early design stage, we introduce a cycle-accurate NoC-based DNN simulator, called DNNoC-sim. To support various operations such as convolution and pooling in the modern DNN models, we first propose a DNN flattening technique to convert diverse DNN operation into MAC-like operations. In addition, we propose a DNN slicing method to evaluate the large-scale DNN models on a resource-constraint NoC platform. The evaluation results show a significant reduction in the off-chip memory accesses compared to the state-of-the-art DNN model. We also analyze the performance and discuss the trade-off between different design parameters.     

A hybrid packet-circuit switched router for optical network on chip

The increasing number of Intellectual Property (IP) cores challenges the traditional electrical Network on Chip (NoC). Silicon nanophotonics becomes a leading technology because of offering several benefits for NoC. Also, On-chip services, including guaranteed service and best-effort service, have different traffic characteristics. This has an important influence on the performance of NoC. This paper proposes a hybrid packet-circuit switched router for optical Network on Chip (ONoC). It can support optical circuit switching (OCS) and optical packet switching (OPS) in parallel and simultaneously, in order to optimize the performance of the network with both services. According to the simulation results, the proposed architecture achieves lower latency and higher throughput than the traditional architectures in the same network scale.     

基于深度学习的文本分类技术研究进展

随着深度学习技术的快速发展,许多研究者尝试利用深度学习来解决文本分类问题,特别是在卷积神经网络和循环神经网络方面,出现了许多新颖且有效的分类方法。对基于深度神经网络的文本分类问题进行分析,介绍卷积神经网络、循环神经网络、注意力机制等方法在文本分类中的应用和发展,分析多种典型分类方法的特点和性能,从准确率和运行时间方面对基础网络结构进行比较,表明深度神经网络较传统机器学习方法在用于文本分类时更具优势,其中卷积神经网络具有优秀的分类性能和泛化能力。在此基础上,指出当前深度文本分类模型存在的不足,并对未来的研究方向进行展望。     

动态层次小组多播路由的研究

在分析现有多播路由协议的基础上,本文提出了动态层次小组多播路由。多播数据源可以根据自治系统（AS）中接收者的多少而动态决定发送多播数据的方式。在一个AS中,对于一个特定的多播源,当多播数据接收者的数量较少时,用单播的方式;当自治系统中接收者较多时,用组地址的方式转发多播数据。同样,AS中的分枝节点亦可以根据该策略构造以分枝节点为根的动态管理方式,从而动态地构造层次多播路由体系结构。本文还定义了动态层次多播路由算法的数据结构以及功能模块,并从网络带宽与路由器计算两方面面分析了动态层次小组多播路由协议分发多播数据时的代价。     

Multicast support in multi-chip centralized schedulers in Input Queued switches

IQ switches store packets at input ports to avoid the memory speedup required by OQ switches. However, packet schedulers are needed to determine an I/O (input/output) interconnection pattern that avoids conflicts among packets at output ports. Today, centralized, single-chip, scheduler implementation are largely dominant. In the near future, the multi-chip scheduler implementation will be needed to reduce the hardware scheduler complexity in very large, high-speed, switches. However, the multi-chip implementation implies introducing a non-negligible delay among input and output selectors used to determine the I/O interconnection pattern at each time slot. This delay, mainly due to inter-chip latency, requires modifications to traditional scheduling algorithms, which normally rely on the hypothesis that information exchange among selectors can be performed with negligible delay. We propose a novel multicast scheduler, named IMRR, an extension of a previously proposed multicast scheduling algorithm named mRRM, making it suitable to a multi-chip implementation, and examine its performance by simulation.     

Multicast performance measurement on a high-performance IP backbone

This paper describes an approach to IP multicast performance measurement on the National Science Foundation's very-high-speed Backbone Network Service (vBNS). Using OC-12c ATM attached workstations that act as either multicast senders or receivers distributed throughout the vBNS backbone, we create arbitrary topologies, generate synthetic IP multicast traffic and measure loss encountered. We present packet loss results as a function of time and as a function of router hop count (distribution tree depth). We analyze the multicast join latency incurred and consider both raw results and results with the loss due to the initiation of multicast state removed. We then correlate these loss measurements to the multicast distribution tree where results show less correlation than expected. We attribute various factors including unbalanced non-binary distribution trees, loss on the receiver links and the individual characteristics of different wide area links as factors affecting the correlation.     

基于集成FEC的可靠多目广播及性能评价

对点对点通信，数据的传输常使用自动重复求技术，而前向纠错多用于半可靠实时传输。然而，ＡＲＱ用于多目广播的性能不佳。     

NISHA: A fault-tolerant NoC router enabling deadlock-free Interconnection of Subnets in Hierarchical Architectures

Decrease in the Integrated Circuit (IC) feature sizes leads to the increase in the susceptibility to transient and permanent errors. The growing rate of such errors in ICs intensifies the need for a wide range of solutions addressing reliability at various levels of abstractions. Network on Chip (NoC) architecture has been introduced to address the increasing demand for communication bandwidth among processing cores. The structural redundancy inherited in NoC-based system can be leveraged to improve reliability and compensate for the effects of failures. In this paper, we propose a fault-tolerant NoC router NISHA, which stands for No-deadlock Interconnection of Subnets in Hierarchical Architectures. Armed with a new flow control mechanism, as well as an enhanced Virtual Channel (VC) regulator, the proposed router can mitigate the effects of both transient and permanent errors. A Dynamic/Static virtual channel allocation with respect to the local and global traffic is supported in NISHA; thereby, it maintains a deadlock-free state in the presence of routers or link failures in hierarchical topologies. Experimental results show an enhanced operation of NoC applications as well as the decrease in the average latency and energy consumption.     

An Efficient Tree-Based Multicasting Algorithm on Wormhole-Routed Star Graph Interconnection Networks Embedded with Hamiltonian Path

Multicasting is an important issue for numerous applications in parallel and distributed computing. In multicasting, the same message is delivered from a source node to an arbitrary number of destination nodes. The star graph interconnection network has been recognized as an attractive alternative to the popular hypercube network. In this paper, we propose an efficient and deadlock-free tree-based multi-cast routing scheme for wormhole-routed star graph networks with hamiltonian path. In our proposed routing scheme, the router is with the input-buffer-based asynchronous replication mechanism that requires extra hardware cost. Meanwhile, the router simultaneously sends incoming flits on more than one outgoing channel. We perform simulation experiments with the network latency and the network traffic. Experimental results show that the proposed scheme reduces multicast latency more efficiently than other schemes.     

结合概率图模型与DNN的DDoS攻击检测方法

从传统网络到物联网,分布式拒绝服务攻击一直是网络安全的隐患。为提高分布式拒绝服务攻击的检测率,提出基于概率图模型与深度神经网络的DDoS攻击检测方案。该检测方案由数据预处理阶段和攻击检测阶段组成,在数据预处理阶段,研究了正常数据包与攻击包的区别,分别从TCP、UDP以及IP数据包包头信息提取出较高维的统计特征,根据随机森林计算的特征重要性因子,保留了前22个特征用于流量检测。22个统计特征通过概率图模型的隐马尔科夫算法进行聚类,然后将聚类结果通过检测阶段的深度神经网络对网络数据进行进一步的检测。在CICDoS数据集上进行验证性实验,结果表明,该检测方法的准确率最高可达99.35%,最低检测误报率和漏警率分别可达0.51%和0.12%。     

Online clinical decision support system using optimal deep neural networks

The Internet of Health things (IoHT) has numerous applications in healthcare by integrating health monitoring things like sensors and medical devices for remotely observe patient’s records to provide smarter and intelligent medicare services. To avail best healthcare services to the users using the e-health applications, in this paper, we propose an IoT with cloud based clinical decision support system for the prediction and observance of Chronic Kidney Disease (CKD) with its level of severity. The proposed framework collects the patient data using the IoT devices attached to the user which will be stored in the cloud along with the related medical records from the UCI repository. Furthermore, we employ a Deep Neural Network (DNN) classifier for the prediction of CKD and its level of severity. A Particle Swarm Optimization (PSO) based feature selection method is also used to improve the performance of DNN classifier. The proposed model is validated by employing the benchmark CKD dataset. Different classifiers are employed to compare the performance of the proposed model under several classification measures. The proposed DNN classifier alone predicts CKD with an accuracy of 98.25% and is further enhanced to 99.25 by PSO-FS method. At the same time, the improved classification performance is verified with higher values of 98.03 specificity, 99.25 accuracy, 99.39 F-score and 98.40 kappa value respectively.     

A novel minimum delay maximum flow multicast algorithm to construct a multicast tree in wireless mesh networks

The construction of multicast tree within given constraints, such as delay and capacity, is becoming a major problem in many wireless networks, especially wireless mesh networks (WMN). Due to the limited capacity of the wireless node, a multicast call may be dropped if there is no multicast tree formed within the given constraints. In this paper, we propose a new multicast tree construction algorithm which has maximum traffic flow and minimum delay under capacity constraints. The problem of multicast is formulated as a Linear Programming (LP) problem with associated constraints. A cost function (CF) is defined to choose the less loaded route among the available ones. A Minimum Delay Maximum Flow Multicast (MDMF) algorithm is proposed to solve this problem using CF and associated constraints. The performance of the proposed algorithm and CF is evaluated and compared with well-known algorithms with respect to packet delivery fraction, latency, and network throughput. The results obtained show that the proposed algorithm has a lesser number of transmissions for a given CF. Moreover, the proposed algorithm has high throughput, packet delivery fraction and less latency compared to other well-known algorithms in this category.     

长期演进切换中的TCP性能改进

针对长期演进(LTE)网络中切换引起的TCP数据包乱序问题,提出了一种动态超时重传计时器(RTO)算法——DRTO。DRTO算法的核心是利用TCP数据包的序号来区分新旧数据包,通过新旧数据包序号的差值来取代以往计算传统RTO时很难确定的乘积因子。该算法不需要对切换机制进行修改,就可以解决用户切换完成前接收的数据包(源基站接续转发的数据包)与切换完成后接收的数据包(服务器发送的数据包)的乱序问题。最后,在NS-2仿真平台下,分析比较DRTO算法与传统RTO算法。仿真结果表明,在吞吐量、重传数据包个数和时延三个性能指标上,DRTO算法均优于传统RTO算法。