We propose a scalable graph processing hardware accelerator called NOVA that is based on a novel vertex management architecture that decouples the execution of reduction and propagation operations in the popular vertex-centric graph processing paradigm. This allows us to store the working set in off-chip memory and utilize the available on-chip memory as a buffer to hide the latency of DRAM accesses instead of a traditional cache. This overcomes one of the key drawbacks of almost all the prior works which require temporal partitioning of graphs to scale to large graphs. We develop a cycle-accurate model of the architecture in gem 5 and demonstrate that NOVA exhibits near-perfect weak and strong scaling while scaling to large graphs by spatially tiling multiple nodes. In addition, our simulations show that NOVA is $2.35 \times$ better than a state-of-the-art graph accelerator (PolyGraph) while using a fraction of the on-chip memory on a synthetic graph with 134M vertices and over 2.14B edges.

Deep reinforcement learning (DRL) enables autonomic optical networking by allowing the network control and management systems to self-learn successful networking policies from operational experiences. This paper proposes a transfer learning approach for effective and scalable DRL in optical networks. We first present a modular DRL agent design to facilitate retrieving and transferring relevant knowledge between tasks requiring different dimensions of network state data. In particular, we partition network state data into common states, which contain generic information critical to multiple tasks (e.g., the spectrum utilization on fiber links), and task-specific states that are only used by a specific task (e.g., the utilization of virtual network functions). Separate neural network blocks are employed to process different state data. Based on the modular agent design, a multi-task learning (MTL) aided knowledge transferring scheme is proposed. The proposed scheme trains an MTL agent that can master multiple tasks simultaneously and thus enables to learn and transfer better-generalized knowledge across tasks. We perform case studies on the proposed transfer DRL approach considering two scenarios, namely, knowledge transferring between routing, modulation and spectrum assignment (RMSA) tasks for different networks and knowledge transferring from RMSA tasks to anycast service provisioning tasks. The DRL designs for RMSA and anycast service provisioning, including the state, action, and reward formulations and the training mechanisms, are also elaborated. Performance evaluations under both scenarios show that the proposed approach can effectively expedite the training processes of the target tasks and improve the ultimate service throughput.

Accurate delay estimation is one of the enablers of future network connectivity services, as it facilitates the application layer to anticipate network performance. If such connectivity services require isolation (slicing), such delay estimation should not be limited to a maximum value defined in the Service Level Agreement, but to a finer-grained description of the expected delay in the form of, e.g., a continuous function of the load. Obtaining accurate end-to-end (e2e) delay modeling is even more challenging in a multi-operator (Multi-AS) scenario, where the provisioning of e2e connectivity services is provided across heterogeneous multi-operator (Multi-AS or just domains) networks. In this work, we propose a collaborative environment, where each domain Software Defined Networking (SDN) controller models intra-domain delay components of inter-domain paths and share those models with a broker system providing the e2e connectivity services. The broker, in turn, models the delay of inter-domain links based on e2e monitoring and the received intra-domain models. Exhaustive simulation results show that composing e2e models as the summation of intra-domain network and inter-domain link delay models provides many benefits and increasing performance over the models obtained from e2e measurements.

While the Fat-Tree network topology represents the dominant state-of-art solution for large-scale HPC networks, its scalability in terms of power, latency, complexity, and cost is significantly challenged by the ever-increasing communication bandwidth among tens of thousands of heterogeneous computing nodes. We propose 3D-Hyper-FleX-LION, a flat hybrid electronic-photonic interconnect network that leverages the multichannel nature of modern multi-terabit switch ASICs (with 100 Gb/s granularity) and a reconfigurable all-to-all photonic fabric called Flex-LIONS. Compared to a Fat-Tree network interconnecting the same number of nodes and with the same oversubscription ratio, the proposed 3D-Hyper-FleX-LION offers a 20% smaller diameter, 3x lower power consumption, 10X fewer cable connections, and 4x reduction in the number of transceivers. When bandwidth reconfiguration capabilities of Flex-LIONS are exploited for non-uniform traffic workloads, simulation results indicate that 3D-Hyper-FleX-LION can achieve up to 4x improvement in energy efficiency for synthetic traffic workloads with high locality compared to Fat-Tree.

In this paper, we propose an incentive-driven virtual network function service chaining (VNF-SC) framework for optimizing the cross-stratum resource provisioning in multi-broker orchestrated inter-datacenter elastic optical networks (IDC-EONs). The proposed framework employs a non-cooperative hierarchical game-theoretic mechanism, where the resource brokers and the VNF-SC users play the leader and the follower games, respectively. In the leader game, the brokers calculate VNF-SC service schemes for users and compete for the provisioning tasks. While in the follower game, the users compete for VNF-SC services for jointly optimizing the resource cost and the received quality-of-service. We first elaborate on the modeling of the follower game, discuss the existence of Nash equilibrium and propose a mixed-strategy gaming approach enabled by an auxiliary graph-based algorithm to facilitate users selecting the most appropriate service schemes. Then, under the assumption that the brokers are aware of the principle of the follower game, we present the model for the leader game and develop a time-efficient heuristic algorithm for brokers to compete for the provisioning tasks. Simulations show that the proposed incentive-driven VNF-SC framework significantly improves the network throughput (i.e., $>4.8 \times$ blocking reduction) while assisting users and brokers in achieving higher utilities compared with existing solutions.

Silicon Photonics (SiPs) can overcome the energy and bandwidth limitations of electrical interconnects in networks-on-chip (NoCs) and enable efficient global all-to-all connectivity-the ideal from a performance perspective. Unfortunately, state-of-the-art SiP switching fabrics impose power overheads for thermo-optical control of microring resonators (MRs), excessive crosstalk, or challenging physical layout. The Arrayed Waveguide Grating Router (AWGR) is a SiP device that provides layout-efficient and scalable all-to-all connectivity through wavelength-routing on a passive and compact platform. Recent technological advances now enable AWGR integration with significantly reduced footprint (2), crosstalk (<;-38dB), and loss (<;2dB), making AWGRs emerge as a major enabler for energy-efficient all-to-all connectivity in NoCs. This paper, for the first time, populates the design space of AWGR-based NoC topologies and compares AWGRs to state-of-the-art SiP fabrics and aggressive electrical baselines. Our results make a compelling case for AWGRs in interposer-based systems with processor disintegration, which have high bandwidth demands, large on-chip distances, and less stringent area constraints. AWGR-based NoCs can offer an average of 1.67× power savings, 1.23× speed-up, and 1.5× lower energy-delay-product (EDP) on PARSEC3.0/SPLASH-2× workloads compared to electrical baselines, and fewer waveguides, wavelengths, or MRs than alternative SiP crossbar fabrics.

In future performance improvement of the basic building block of supercomputers has to come through increased integration enabled by 3D (vertical) and 2.5D (horizontal) die-stacking. But to take advantage of this integration we need an interconnection network between the memory and compute die that not only can provide an order of magnitude higher bandwidth but also consume an order of magnitude less power than today's state of the art electronic interconnects. We show how Arrayed Waveguide Grating Router-based photonic interconnects implemented on the silicon interposer can be used to realize a 16 × 16 photonic Network-on-Chip (NoC) with a bisection bandwidth of 16 Tb/s. We propose a baseline network, which consumes 2.57 pJ/bit assuming 100% utilization. We show that the power is dominated by the electro-optical interface of the transmitter, which can be reduced by a more aggressive design that improves the energy per bit to 0.454 pJ/bit at 100% utilization. Compared to recently proposed interposer-based electrical NoC's we show an average performance improvement of 25% on the PARSEC benchmark suite on a 64-core system using the Gem5 simulation framework.

This article proposes a scalable wavelength-routed optical Network on Chip (NoC) based on the Spidergon topology, named Power-efficient Scalable Wavelength-routed Network-on-chip (PeSWaN). The key idea of the proposed all-optical architecture is the utilization of per-receiver wavelengths in the data network to prevent network contention and the adoption of per-sender wavelengths in the control network to avoid end-point contention. By performing a series of simulations, we study the efficiency of the proposed architecture, its power and energy consumption, and the data transmission delay. Moreover, we compare the proposed architecture with electrical NoCs and alternative ONoC architectures under various traffic patterns.

This work studies constrained rate maximization problem for clustered device-to-device (D2D) communication in cellular networks. In the cluster, only one capable user equipment (cluster head) downloads data from the base station. D2D communication is used for local data sharing between the cluster head and other user equipments in the cluster. When sharing and reusing cellular spectrum resources, D2D links must control the potential impact of mutual interference. In a scenario that D2D is only allowed by eNB when the cochannel cellular QoS constraint can be achieved, we investigate joint precoding strategy of D2D and cochannel cellular transmission subject to the cellular outage probability constraint. Based on estimated cellular channel state information (CSI), we analyze the feasibility and optimality of the constrained rate maximization problem and provide optimized solutions for different scenarios. Our simulation demonstrates the advantage of the proposed precoding strategy in satisfying cellular QoS and improving D2D rate.

Elastic optical networking (EON) has emerged in recent years as a promising solution for implementing flexible bandwidth channels (flexpaths) that efficiently match the allocated bandwidth with the traffic demand using agile granularities of spectrum allocation. However, the additional flexibility in such networks raises challenges in terms of efficient control and management of spectrum resources. Among them, three important issues are (1) mitigation of spectral fragmentation, (2) implementation of impairment awareness and enhancement of robustness against impairments for potentially large-bandwidth flexpaths, and (3) design of an efficient restoration scheme to combat network failures. This paper presents an adaptive spectrum control and management scheme, which includes: dynamic on-demand spectral defragmentation, adaptive combinational quality of transmission (QoT) restoration (ACQR) and supervisory channel-assisted active restoration, to account for the three issues above. We present scalable networking algorithms and experimental demonstrations that address these issues in an EON testbed. Simulation results show that the defragmentation technique is capable of reducing the provisioning blocking probability by half with only one defragmentation module on each link. Then, we also show that the ACQR can efficiently restore many degraded flexpaths on the same impaired link while reducing the restoration blocking probability by a factor of 10 compared with the conventional rerouting method. At last, we show via simulation the advantages of using supervisory channels to determine restoration path quality and selection in EON restorations. This paper also presents experimental demonstrations to corroborate the effectiveness and feasibility of implementing these capabilities in next generation optical networks.

This paper discusses the architecture and performance studies of Datacenter Optical Switch (DOS) designed for scalable and high-throughput interconnections within a data center. DOS exploits wavelength routing characteristics of a switch fabric based on an Arrayed Waveguide Grating Router (AWGR) that allows contention resolution in the wavelength domain. Simulation results indicate that DOS exhibits lower latency and higher throughput even at high input loads compared with electronic switches or previously proposed optical switch architectures such as OSMOSIS [4, 5] and Data Vortex [6, 7]. Such characteristics, together with very high port count on a single switch fabric make DOS attractive for data center applications where the traffic patterns are known to be bursty with high temporary peaks [13]. DOS exploits the unique characteristics of the AWGR fabric to reduce the delay and complexity of arbitration. We present a detailed analysis of DOS using a cycle-accurate network simulator. The results show that the latency of DOS is almost independent of the number of input ports and does not saturate even at very high (approx 90%) input load. Furthermore, we show that even with 2 to 4 wavelengths, the performance of DOS is significantly better than an electrical switch network based on state-of-the-art flattened butterfly topology.

This paper proposes a Computational Intelligence framework for network security management. The framework uses reinforcement learning and Bayesian methods to achieve cross-layer optimization in heterogeneous, multi-layer wireless and wireline networks. Metrics based on the reputation of a node are used to measure performance. OPNET simulations results indicate that routing algorithms based on the newly proposed framework exhibit nearly four times improved network performance measured in terms of goodput.

Traditional hop-by-hop dynamic routing makes inefficient use of network resources as it forwards packets along already congested shortest paths while uncongested longer paths may be underutilized. To maintain network-wide load balancing, we propose Autonomous Network management with Team learning based Self-configuration (ANTS) which attempts to manage a feasible route for traffic flow with QoS constraints in heterogeneous networks. To enable cognitive intelligence for network-wide load balancing, we implement a cross-layer mechanism in which learning agents in middleware layer can monitor the queue sizes of MAC layer, thereby allowing for the discovery of optimal routes. We present OPNET simulation results illustrating that, in comparison to original OSPF and AODV (2.18 Mbits/s with 46.46% packet loss rate), ANTS dramatically achieves a higher packet delivery (9.57 Mbits/s with 0.53% packet loss rate).

In distributed wireless access networks, the short-term unfairness of IEEE 802.11 distributed coordination function (DCF) has been revealed by many works. In this paper, a modified DCF, based on the principle of priority collision resolution (PCR), is proposed to improve the short-term fairness in distributed access wireless networks. Our PCR-DCF achieves short-term fairness improvement without estimating behaviors of other users and the system contention level, such as the number of active users. Only the capability to identify collision is required. Theoretical analysis is carried out to validate the performance of PCR-DCF based on a slotted system model. Both simulations and analyses show that PCR-DCF has a much smaller packet transmission delay jitter than IEEE 802.11 DCF, with little degradation on the average packet transmission delay. Moreover, the analysis and simulation results indicate that PCR- DCF also experiences a much lower packet drop rate than IEEE 802.11 DCF in almost all load ranges with reasonable retransmission limit.

We present the design and evaluation of 14, a network infrastructure that enables information exchange and collaboration among different domains. 14 can help with network management in many scenarios, such as when eliminating the unwanted traffic to improve the network performance as well as diagnosing the network problems. We present the distributed denial-of-service (DDoS) attack as an example to demonstrate the advantages of 14. Simulation results show that 14 can significantly reduce the amount of DDoS attack packets and dramatically improve the quality of services received by legitimate users. Our design provides attractive properties, such as incremental deployment as well as incentives for such deployment etc.

This paper presents the use of space-time coding techniques to fully exploit polarization diversity in coherent optical systems without the need for polarization tracking. Our approach is analogous to a situation of two transmit antennas and two receive antennas in a wireless communication system. Utilizing the two polarizations intelligently, the highly effective technique of orthogonal space-time block coding is applied to combat polarization uncertainty and signal losses due to free- space non-idealities. The use of space-time block coding and differential space-time block coding are modeled and tested successfully. Significant performance gain is demonstrated by our new coherent optical multiple-input multiple-output (COMIMO) system over many traditional coherent optical communication techniques.

In this paper, we propose a spectral-slice scheme and a chip-level detection for the TTFR-based asynchronous SPECTS-OCDMA system combined with time-encoding. The spectral-slice scheme can reduce the interference probability because the MAI with the different frequency has no adverse effects on the channel collision rate. Moreover, the chip-level detection can enhance the tolerance of MAI because the channel collision does not occur as long as either of the detection time positions receives no MAI. As a result, these techniques mitigate channel collisions. We analyze the channel collision rate theoretically, and show that the proposed system can achieve a lower channel collision rate in comparison with the conventional systems.

In a multi-source video streaming system, premature draining of low-power nodes can cause sudden failures of peer connections and degrade streaming performance. To solve this problem, we propose an energy-aware scheduling (EAS) scheme to better distribute the streaming load among different peers by jointly considering network conditions and node energy levels. We model the proposed scheme using a rate/energy-distortion optimization framework and heuristically solve it using the concept of asynchronous clocks. Simulation studies show that the proposed EAS scheme can achieve comparable streaming quality while consuming less energy.

This paper proposes to leverage multi-source multi-path diversity to design a video streaming system for supporting concurrent video-on-demand (VoD) services over wireless mesh networks (WMNs). By integrating a wireless interference model into consideration, we have a more realistic network model to capture the characteristics of wireless networks. Based on that, we mathematically formulate the route selection problem for the proposed streaming system using rate/interference-distortion optimization framework, and rely on a genetic algorithm to solve it heuristically. Simulation results show that the proposed system has better performance than systems using single-source, single-path and systems that do not consider interference.

This paper investigates code assignment and transmission scheduling schemes for code-limited optical code division multiple access (O-CDMA) networks, where the number of nodes exceeds the number of codes. We present an architectural design of a coordinated access scheme based on a centralized gated-polling controller, and evaluate its performance in terms of scheduling delays and achievable throughputs. In particular, this paper proposes a list-scheduling algorithm for code assignment and analyzes its near-optimal performance. With theoretic analyses and simulations, we derive the relationships among system capacity, traffic load, and network performance, and establish an analytical framework to address practical issues regarding O-CDMA network planning. Performance results indicate that the code-limited system can achieve high throughputs at the expense of scheduling delays. The capacity requirement of the control channel is also investigated to ensure system stability. These results provide an in-depth insight into network dynamics of code-limited systems and can serve as guidelines for network planning and algorithm design.

Customer retrial is a common feature in practical service systems. This paper develops an analytical approach that incorporates reduced load analyses employing a retrial model, and investigates the effect of request retrials on the performance of wavelength-routed optical networks. The analytical approach can evaluate network performance in terms of request blocking probability, request loss probability, and total carried traffic. We analyze a representative network (NSFnet) with different retrial features and validate the proposed approach with simulation experiments. Analytical results indicate that the retrials increase the blocking rates of fresh request while decreasing the request loss rate, thus leading to a high total carried traffic in a network. For the routing schemes used in this work, the analytical results set a lower bound to the request loss probability and an upper bound to the total available traffic. The developed model allows performance analyses and service optimisation including customer retrials.

This paper describes a novel sequential wavelength-time-space (sWTS) scheduling algorithm to solve the arbitration problem in an all-optical packet switch router and shows the simulation as well as the hardware implementation results. The simulation results with self-similar traffic input demonstrate that the sWTS arbitration algorithm effectively improves the packet blocking rate, forwarding latency and jitter as the number of wavelength channels per port and the number of recirculation buffer ports in the switch fabric increase. The proposed algorithm further facilitates a very fast and hardware-efficient mixed-tree output port arbiter design. The hardware implementation is evaluated in terms of area and delay for various switch sizes with both the Xilinx XCV1000E FPGA and a 0.25-micron commercial ASIC library.

This paper proposes a generalized approach for throughput analysis and demonstrates its application for performance evaluation in OCDMA networks. The proposed analysis approach takes into account not only the characteristics of the physical layer to support arbitrary BER models, but also the nodal architecture design at higher layers. In particular, we investigate the effects of packet corruption, channel collision, and destination contention on network performance, and obtain steady-state estimation for effective throughput. This throughput estimation sets a tight upper bound for achievable throughput in OCDMA networks with random access schemes. The proposed analytic approach leads to results consistent with simulation experiments. We demonstrate the effectiveness of this approach by analyzing a representative OCDMA scheme with various system parameters. The derived performance metrics have proven effective in characterizing the OCDMA network dynamics.

This paper proposes and analyzes a two-stage optical switch architecture supporting asynchronous variable length packet switching for optical label switching network applications. The proposed scheduling algorithm combines contention resolution in wavelength, time, and space domains with the input channel look-ahead scheduling scheme and the output channel load-balance scheme. The network-wide simulation results with self-similar traffic show that the new architecture and scheduling algorithm lead to a uniform output wavelength channel utilization and effectively reduce the packet loss rate and the end-to-end jitter compared with the single-stage switch at the expense of an extra look-ahead delay.

This paper presents an integrated network management system (INMS) for optical-label switching networks. The INMS addresses the issues of managing multi-layer heterogeneous networks in an integrated way; and it provides a platform for integrated multi-layer (IP/WDM) traffic engineering, fault management, and restoration. We demonstrate a prototype implementation of the INMS, which performs on-line traffic engineering over label switching and optical circuit switching layers with integrated intelligence.

This paper proposes a joint server/peer video streaming architecture for wireless networks, where a receiver can access a video server via an access point using the infrastructure mode and at the same lime communicate with its peers using the ad hoc mode of its IEEE 802.11 interface card. We introduce a joint infrastructure/peer-to-peer, receiver-driven streaming scheme, and formulate it as a combinatorial optimization problem. We decouple the problem into two steps: first selecting the sender (server or peer) by introducing asynchronous clocks, and then applying point-to-point rate-distortion optimization algorithm between a specific sender-receiver pair. Simulation results show that our joint approach has better performance than those systems with single server or with round-robin selection scheme.

Contention resolution schemes are key determinant of network performances in optical packet switched networks. This paper presents a contention resolution scheme in wavelength and space dimensions. By integrating deflection routing with wavelength conversion, this paper investigates their effects on packet-loss performance through simulation studies. In particular, we propose a load-sensitive deflection routing algorithm to compute preferred and deflection routing tables. The proposed algorithm proves effective in resolving contention and in reducing packet-loss rates. The load-sensitive deflection routing can share network resources between multiple nodes effectively and produce the same effect as increasing the number of wavelengths in each node. Simulation comparisons with other deflection routing schemes indicate that the load-sensitive deflection routing is the most effective when the load is high in some parts of a network.

This paper discusses the architecture, protocol, analysis, and experimentation of optical packet switching routers incorporating optical-label switching (OLS) technologies and electronic edge routers with traffic shaping capabilities. The core optical router incorporates all-optical switching with contention resolution in wavelength, time, and space domains. It is also capable of accommodating traffic of any protocol and format, and supports packet, flow, burst, and circuit traffic. The edge router is designed to achieve traffic shaping with consideration for quality of service and priority based class-of-service. Simulation results show packet loss rates below 0.3% at load 0.7 and jitter values below 18 μs. The traffic shaping reduces the packet loss rate by a factor of ∼5 while adding negligible additional latency. The OLS core routers and the electronic edge routers are constructed including the field-programmable-gate-arrays incorporating the wavelength-aware forwarding and contention resolution algorithms. The experiment shows optical-label-based packet switching with a packet loss rate near 0.2%.

Traffic shaping at the edge router is one possible solution to mitigate the adverse effects of self-similar traffic in optical packet-switched networks. In this paper we propose an optical packet assembly mechanism and study its impacts on the traffic characteristics. In addition to the Hurst parameter, we apply new measures to describe the short-term correlation and to investigate the effect of the system parameters in the assembly mechanism. Our simulation study indicates that this assembly mechanism can reduce the traffic correlation and the degree of self-similarity at an acceptable latency in terms of the assembly delay.