Search

WOS: 000526412600010, This study proposes a distributed cluster-based routing algorithm, distributed clustering with decreased uncovered nodes (DiCDU), alleviating the uncovered node problem occurring after the election of cluster heads. For the purpose of increasing the performance of the system in terms of the lifetime of the network, an energy-based cluster head election without requiring any extra control message is used. Moreover, an energy saving mode is proposed, in which the nodes perform static clustering controlled by predefined thresholds to conserve batteries. Extensive simulations are conducted for different values of cluster radius and the number of nodes and proposed schema is compared with similar methods, including a time-based cluster-head selection algorithm for low-energy adaptive clustering hierarchy, a clustering algorithm based on time delay, and regional energy aware clustering with the isolated nodes, for the death of the first, the half and the last nodes with the number of uncovered nodes. The results show that DiCDU outperforms compared approaches for all values of cluster radius and the number of nodes.

For the operation of the spinning frame in the spinning mill, Hall sensors are used to test the running rate of each axis of the spinning frame. In order to centrally monitor the operation of the spinning frame, analyse the production site environment, communication distance and node layout, ZigBee is used to establish a networked monitoring system to improve the stability of rate acquisition and enhance the stability of data transmission. According to the actual condition at the production site, the ZigBee traditional routing algorithm is improved to be more convenient and efficient in the textile mill environment and have the packet loss rate reduced. The rate collection node sends the collected data to the coordinator. The coordinator uploads the data to the upper computer by WI-FI module through WI-FI gateway. The results show that the improved routing algorithm greatly reduces the packet loss rate and the entire monitoring system is stable and reliable.

The digital transformation is pushing the existing network technologies towards new horizons, enabling new applications (e.g., vehicular networks). As a result, the networking community has seen a noticeable increase in the requirements of emerging network applications. One main open challenge is the need to accommodate control systems to highly dynamic network scenarios. Nowadays, existing network optimization technologies do not meet the needed requirements to effectively operate in real time. Some of them are based on hand-crafted heuristics with limited performance and adaptability, while some technologies use optimizers which are often too time-consuming. Recent advances in Deep Reinforcement Learning (DRL) have shown a dramatic improvement in decision-making and automated control problems. Consequently, DRL represents a promising technique to efficiently solve a variety of relevant network optimization problems, such as online routing. In this paper, we explore the use of state-of-the-art DRL technologies for real-time routing optimization and outline some relevant open challenges to achieve production-ready DRL-based solutions., 6 pages, 4 figures

Various types of communication networks are constantly emerging to improve the connectivity services and facilitate the interconnection of various types of devices. This involves the development of several technologies, such as device-to-device communications, wireless sensor networks and vehicular communications. The various services provided have heterogeneous requirements on the quality metrics such as throughput, end-to-end latency and jitter. Furthermore, different network technologies have inherently heterogeneous restrictions on resources, e.g., power, interference management requirements, computational capabilities, etc. As a result, different network operations such as spectrum management, routing, power control and offloading need to be performed differently. Mathematical optimization techniques have always been at the heart of such design problems to formulate and propose computationally efficient solution algorithms. One of the existing powerful techniques of mathematical optimization is Benders decomposition (BD), which is the focus of this article. Here, we briefly review different BD variants that have been applied in various existing network types and different design problems. These main variants are the classical, the combinatorial, the multi-stage, and the generalized BD. We discuss compelling BD applications for various network types including heterogeneous cellular networks, infrastructure wired wide area networks, smart grids, wireless sensor networks, and wireless local area networks. Mainly, our goal is to assist the readers in refining the motivation, problem formulation, and methodology of this powerful optimization technique in the context of future networks. We also discuss the BD challenges and the prospective ways these can be addressed when applied to communication networks' design problems., Accepted in IEEE Network Magazine, Aug. 2019, Article No.: NETWORK-19-00414.R1

In this study the authors are interested in safety-critical real-time applications implemented on distributed architectures supporting the time-sensitive networking (TSN) standard. The on-going standardisation of TSN is an IEEE effort to bring deterministic real-time capabilities into the IEEE 802.1 Ethernet standard supporting safety-critical systems and guaranteed quality-of-service. TSN will support time-triggered (TT) communication based on schedule tables, audio-video-bridging (AVB) flows with bounded end-to-end latency as well as best-effort messages. The authors first present a survey of research related to the optimisation of distributed cyber-physical systems using real-time Ethernet for communication. Then, the authors formulate two novel optimisation problems related to the scheduling and routing of TT and AVB traffic in TSN. Thus, the authors consider that they know the topology of the network as well as the set of TT and AVB flows. The authors are interested to determine the routing of both TT and AVB flows as well as the scheduling of the TT flows such that all frames are schedulable and the AVB worst-case end-to-end delay is minimised. The authors have proposed an integer linear programming formulation for the scheduling problem and a greedy randomised adaptive search procedure-based heuristic for the routing problem. The proposed approaches have been evaluated using several test cases.

The rise of Software Defined Networking (SDN) presents an opportunity to overcome the limitations of rigid and static traditional Internet architecture and provide services like network layer multicast for live video streaming. In this paper we propose mCast, an SDN-based architecture for live streaming, to reduce the utilization of network and system resources for both Internet Service Providers (ISP) and Content Delivery Networks (CDN) by using multicast over the Internet. We propose a communication framework between ISPs and CDNs to enable mCast while retaining user and data privacy. mCast is transparent to the clients and maintains the control of CDNs on user sessions. We developed a testbed and performed large scale evaluation and comparison. Results showed that mCast can improve the video quality received by clients and, for CDNs and ISPs in comparison to IP unicast, mCast can decrease link utilization by more than 50% and network losses to 0%.

Deep reinforcement learning (DRL) has recently revolutionized the resolution of decision-making and automated control problems. In the context of networking, there is a growing trend in the research community to apply DRL algorithms to optimization problems such as routing. However, existing proposals fail to achieve good results, often under-performing traditional routing techniques. We argue that the reason behind this poor performance is that they use straightforward representations of networks. In this paper, we propose a DRL-based solution for routing in optical transport networks (OTNs). Contrary to previous works, we propose a more elaborate representation of the network state that reduces the level of knowledge abstraction required for DRL agents and easily captures the singularities of network topologies. Our evaluation results show that using our novel representation, DRL agents achieve better performance and learn how to route traffic in OTNs significantly faster compared to state-of-the-art representations. Additionally, we reverse engineered the routing strategy learned by our DRL agent, and as a result, we found a routing algorithm that outperforms well-known traditional routing heuristics.

With the proliferation of cloud computing and the expected requirements of future Internet of Things (IoT) and 5G network scenarios, more efficient and scalable Data Centers (DCs) will be required, offering very large pools of computational resources and storage capacity cost-effectively. Looking at todays' commercial DCs, they tend to rely on well-defined leaf-spine Data Center Network (DCN) topologies that not only offer low latency and high bisectional bandwidth, but also enhanced reliability against multiple failures. However, routing and forwarding solutions in such DCNs are typically based on IP, thus suffering from its limited routing scalability. In this work, we quantitatively evaluate the benefits that the Recursive InterNetwork Architecture (RINA) can bring into commercial DCNs. To this goal, we propose rule-based topological routing and forwarding policies tailored to the characteristics of publicly available Google's and Facebook's DCNs. These policies can be programmed in a RINA-enabled environment, enabling fast forwarding decisions in most scenarios with merely neighboring node information. Upon DCN failures, invalid forwarding rules are overwritten by exceptions. Numerical results show that the scalability of our proposal depends on the number of concurrent failures in the DCN rather than its size (e.g., number of nodes/links), dramatically reducing the total amount of routing and forwarding information to be stored at nodes. Furthermore, as routing information is only disseminated upon failures across the DCN, the associated communication cost of our proposals largely outperforms that of the traditional IP-based solutions.

The core-based approach in multipoint communication enhances the solution space in terms of QoS-efficiency of solutions in inter- and intra-domain routing. In an earlier work [A. Karaman, H.S. Hassanein, Extended QoS-framework for Delay-constrained Group Communication, International Journal of Communication Systems, in press.], we showed that the constrained cost minimization solutions in core-based approach proposed to date are restrictive in their search to a subrange of solutions, and we proposed SPAN, a generic framework to process in our identified extended solution space. In this paper, we study the core selection component of SPAN and propose two novel algorithms, SPAN/COST and SPAN/ADJUST, which define the core-selection component of SPAN. SPAN/COST mainly optimizes the cost distances to be traveled between the source-core and core-receiver pairs on the multicast trees, while SPAN/ADJUST selects the cores based on the numbers of nodes they dominate and adjusting the set based on cost. Our algorithms consistently outperform their counterparts proposed to date and can be considered pioneering in their optimization range of multiple metrics and processing in the extended solution space.

Internet protocol (IP) address lookup is one of the major performance bottlenecks in high-end routers. This paper presents an architecture for an IP address lookup engine based on programmable finite-state machines (FSMs). The IP address lookup problem can be translated into the implementation of a large FSM. Our hardware engine is then used to implement this FSM using a structured approach, in which the large FSM is broken down into a set of smaller FSMs which are then mapped into reconfigurable hardware blocks. The design of our hardware engine is based on a regular and well structured architecture, which is easy to scale. Our simulation results demonstrate that the FSM based architecture can easily scale to wire speed performance at OC-192 rates. Unlike previous approaches, the performance of our architecture is not constrained by memory bandwidth and is, therefore, in principle scalable with very large scale integration technology., IEEE

Advances in processor, memory, and radio technology will enable small and cheap nodes capable of sensing, communication, and computation. Networks of such nodes can coordinate to perform distributed sensing of environmental phenomena. In this paper, we explore the directed-diffusion paradigm for such coordination. Directed diffusion is data-centric in that all communication is for named data. All nodes in a directed-diffusion-based network are application aware. This enables diffusion to achieve energy savings by selecting empirically good paths and by caching and processing data in-network (e.g., data aggregation). We explore and evaluate the use of directed diffusion for a simple remote-surveillance sensor network analytically and experimentally. Our evaluation indicates that directed diffusion can achieve significant energy savings and can outperform idealized traditional schemes (e.g., omniscient multicast) under the investigated scenarios.

From the perspective of resource allocation and routing, this paper aims to save as much energy as possible in data center networks. We present a general framework, based on the blocking island paradigm, to try to maximize the network power conservation and minimize sacrifices of network performance and reliability. The bandwidth allocation mechanism together with power-aware routing algorithm achieve a bandwidth guaranteed tighter network. Besides, our fast efficient heuristics for allocating bandwidth enable the system to scale to large sized data centers. The evaluation result shows that up to more than 50% power savings are feasible while guaranteeing network performance and reliability. NPRP grant No.6-718-2-298 from the Qatar National Research Fund. Scopus

We address survivability considerations for telecommunication networks where a part of the network may fail. We focus on single arc failures in multicast networks, with or without network coding. The problem is to compute a routing such that, if any single arc failure occurs, the remaining throughput is as large as possible. In the case of multicast routing without network coding, two ways for computing the impact of an arc failure are considered. Either the throughput of entire Steiner trees containing that arc vanishes, or only the throughput of subtrees, starting from that arc to terminals, is discarded. In the case of multicast with network coding, since the total throughput is computed as the minimum of single max-flow values from the source to any terminal, we consider that a failure of an arc implies that the throughput of all paths containing that arc vanishes. The three obtained models are formulated as linear problems. We solve these problems either directly or by use of column generation. Further, the survivable network coding gain is defined as the ratio of maximal throughput with - over without - network coding, when arc failures can occur. We show that this ratio is unbounded for unitary combination digraphs and hence for any directed networks. We also show that this ratio is equal to one for bidirected graphs with uniform capacities. Finally, some numerical results are carried out on randomly generated networks showing that the gain ratio is equal to one for almost 98 percent of the instances. However, we observe that routing with network coding requires much less redundancy in order to achieve the optimal throughput.

In this work we investigate a new solution dealing with the increase of IP networks energy efficiency. The starting point of this work is Table Lookup Bypass (TLB), a new mechanism to reduce the energy consumption of a linecard overcoming the simple switch-off of network elements: TLB consists in bypassing the lookup operation performed by a router linecard and forwarding the incoming packets on a fixed outgoing interface. The paper defines four different heuristics able to detect a set of TLB operations that can be performed maintaining QoS constraints. The performance evaluation realized on a real network scenario shows that TLB can really improve the energy efficiency of IP network, both during peak hours and off peak hours: in this last case up to 90% of linecards can pass in TLB state.

Nanoscale communication based on Forster Resonance Energy Transfer (FRET) enables nanomachines to communicate with each other using the excited state of the fluorescent molecules as the information conveyer. In this study, FRET-based nanoscale communication is further extended to realize FRET-based nanoscale broadcast communication with one transmitter and many receiver nanomachines, and the performance of the broadcast channel is analyzed information theoretically. Furthermore, an electrically controllable routing mechanism is proposed exploiting the Quantum Confined Stark Effect (QCSE) observed in quantum dots. It is shown that by appropriately selecting the employed molecules on the communicating nanomachines, it is possible to control the route of the information flow by externally applying electric field in FRET-based nanonetworks., Scientific and Technological Research Council of Turkey (TÜBİTAK); Turkish National Academy of Sciences Distinguished Young Scientist Award Program (Turkish Academy of Sciences (TÜBA)-GEBIP); IBM through IBM Faculty Award

This paper presents a new cross layer routing protocol for underwater wireless sensor networks. The solution, termed CARP for Channel-aware Routing Protocol, exploits link quality information for cross layer relay determination. Nodes are selected as relays if they have a (recent) history of successful transmissions to their neighbors. CARP combines link quality with simple topology information (hop count), thus being able to route around connectivity voids and shadow zones. The protocol is also designed to take advantage of power control for selecting robust links. The performance of CARP has been evaluated through ns2-based simulations, and compared to the performance of two previously proposed routing protocols, namely, FBR and DBR. Our results show that CARP robust relay selection mechanism enables it to achieve throughput efficiency that is up to twice the throughput of FBR and almost three times that of DBR. CARP also obtains remarkable performance improvements over FBR and DBR with respect to end-to-end packet latency and energy consumption.

Most research works in optical burst switching (OBS) networks do not take into account the impact of physical layer impairments (PLIs) either by considering fully transparent (i.e., using optical 3R regeneration) or opaque (i.e., electrical 3R regeneration) networks. However, both solutions are not feasible for different reasons. In this paper, we propose a novel translucent OBS network architecture which aims at bridging the gap between the transparent and opaque solutions. In order to evaluate its performance, two different joint regenerator placement and routing heuristics are provided. Simulation results show that our translucent network model achieves performance results as good as those obtained with an opaque solution but with considerably less regenerators.

In those networking scenarios (e.g., the stationary/mobile wireless paradigm) where limited energy storagecapabilities is a limiting factor, power efficiency has been studied extensively for a long time. In the last fewyears, the steadily growing power consumption figures of ICT [1][2] along with a recent public awareness of thepossible environmental effects, broadened the focus of the research efforts towards more power conscioussolutions also for wired networks. One promising solution going in this direction is represented by transparentwavelength division multiplexing (WDM) networks, where all-optical circuits (or lightpaths) carry the networktraffic from the source to the destination node in the optical domain, i.e. without any optical-to-electrical(O-E-O) conversion. To further improve network power efficiency, several approaches have been proposed inthe literature attempting to reduce the power consumption in transparent WDM networks [3][4]. Among them,power aware (PA) routing and wavelength assignment (RWA) algorithms are receiving a lot of attention. Onepossible solution for the PA-RWA problem is to limit the number of devices in the network that need to bepowered-on to support the lightpath requests [5]. These methods, however, have an adverse impact on the lengthof the provisioned lightpaths. In order to efficiently use as much powered-on network resources as possible, thepath length is, on average, longer than the one found with traditional (and less power efficient) shortest pathsolutions. This is in contradiction with the goal of classical RWA algorithms that tend to minimize the resourceusage (i.e., wavelengths and fibers) within the network, in order to minimize the blocking probability. Whenavailability of network resources becomes something that cannot be overlooked, a trade off needs to be assessedbetween power savings and blocking probability.In this talk, a non-conventional solution to the PA-RWA problem is presented. It is based on the intuitionthat, in some cases, relaxing the power minimization constraint can have beneficial effects on the overallnetwork performance, i.e., it can contribute to the reduction of resource fragmentation in the network and, in thisway, lower the blocking probability. The proposed algorithm leverages on a cost function that considers both thepower status of network elements (in this particular case the in-line optical amplifiers on fiber links only) and theinformation about wavelength usage. The algorithm has been tested using a Pan-European core network (i.e.,COST 239). Performance results confirm the presence of a trade-off between energy saving and blockingprobability. They also suggest that a “binary” approach using only the powered-on/powered-off informationmight not always be the best choice. © 2010 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. QC 20120202

As line rates increase, the task of designing high performance architectures with reduced power consumption for the processing of router traffic remains important. In this paper, we present a multi-engine packet classification hardware accelerator, which gives increased performance and reduced power consumption. It follows the basic idea of decision-tree based packet classification algorithms, such as HiCuts and HyperCuts, in which the hyperspace represented by the ruleset is recursively divided into smaller subspaces according to some heuristics. Each classification engine consists of a Trie Traverser which is responsible for finding the leaf node corresponding to the incoming packet, and a Leaf Node Searcher that reports the matching rule in the leaf node. The packet classification engine utilizes the possibility of ultra-wide memory word provided by FPGA block RAM to store the decision tree data structure, in an attempt to reduce the number of memory accesses needed for the classification. Since the clock rate of an individual engine cannot catch up to that of the internal memory, multiple classification engines are used to increase the throughput. The implementations in two different FPGAs show that this architecture can reach a searching speed of 169 million packets per second (mpps) with synthesized ACL, FW and IPC rulesets. Further analysis reveals that compared to state of the art TCAM solutions, a power savings of up to 72% and an increase in throughput of up to 27% can be achieved.

When transport in networks follows the shortest paths, the union of all shortest path trees G union or logical sum SPT can be regarded as the "transport overlay network." Overlay networks such as peer-to-peer networks or virtual private networks can be considered as a subgraph of G union or logical sum SPT. The traffic through the network is examined by the betweenness Bl of links in the overlay G union or logical sum SPT. The strength of disorder can be controlled by, e.g., tuning the extreme value index alpha of the independent and identically distributed polynomial link weights. In the strong disorder limit (alpha-->0), all transport flows over a critical backbone, the minimum spanning tree (MST). We investigate the betweenness distributions of wide classes of trees, such as the MST of those well-known network models and of various real-world complex networks. All these trees with different degree distributions (e.g., uniform, exponential, or power law) are found to possess a power law betweenness distribution Pr[Bl=j] approximately j(-c). The exponent c seems to be positively correlated with the degree variance of the tree and to be insensitive of the size N of a network. In the weak disorder regime, transport in the network traverses many links. We show that a link with smaller link weight tends to carry more traffic. This negative correlation between link weight and betweenness depends on alpha and the structure of the underlying topology.

This paper focuses on the problem of quality of service (QoS) provisioning in optical burst switching (OBS) networks. OBS is a promising photonic network technology aiming at efficient transport of IP traffic by means of statistical multiplexing. The lack of optical memories, however, makes this operation quite complicated. Problems such as unfairness in access to the shared transmission resources, facility in adopting alternative and backup routing, scheduling complexity and so on arise in the conventional OBS architecture. In [1] we proposed the offset-time emulated OBS (E-OBS) architecture, which overcomes all these drawbacks by means of distributed provisioning of the offset time in core nodes. Nonetheless it is still difficult to guarantee a certain level of service quality. Burst preemption mechanism, which, alongside with offset-time differentiation, was proven to be the most effective technique for QoS provisioning in OBS networks. The general drawback of any burst preemptionbased mechanism is that, in case of successful preemption, either the resources reserved for the preempted bursts on outgoing path are wasted or an additional signaling procedure should be carried out in order to release them. In order to avoid wasted resources reservation, in [2] we proposed the Preemption Window (PW) mechanism which enhances the E-OBS for efficient QoS support. In this paper we evaluate exhaustively the performance of the resulting architecture showing all its advantageous with respect to other solutions.

Traffic engineering (TE) describes a domain-wide perspective for traffic control as opposed to application-specific concerns of QoS. MPLS is today's most powerful architecture for TE. Constraint-based routing refers to the process of path assignment to flows to fulfill the TE objective which has been defined in a number of multi-constraint ways, and is the most significant component of TE. In this paper, we review constraint-based routing algorithms. We first provide an overview of MPLS/TE and place constraint-based routing within this approach. We then look into the potential path attributes and the underlying platform to characterize these algorithms before we present our review. Our study provides a comprehensive reference to the operational structure and specifications of constraint-based routing algorithms. Bogazici University IEEE TUBITAK Publisher's Version WOS:000239594100012

We present a router architecture with electronic queuing and a passive optical multi-channel network, which is based upon a single-hop star topology with an arrayed waveguide grating (AWG) at its center. The AWG's property of spatial wavelength reuse and both fixed-tuned and tunable transceivers enable simultaneous control and data transmission. Our proposed medium access control protocol supports traffic differentiation and utilizes EOF (earliest deadline first) to schedule the traffic from input ports to output ports on the router. Our simulations show that the router treats QoS (quality of service) traffic well. ©2005 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.Category number 05EX1035; Code 66090

Submitted on behalf of EDAA (http://www.edaa.com/); International audience; In this paper we introduce a new task model that is specifically targeted towards representing stream processing applications. Examples of such applications are those involved in network packet processing (such as a software-based router) and multimedia processing (such as an MPEG decoder application). Our task model is made up of two parts: (i) a new task structure to accurately model the software structures of stream processing applications such as conditional branches and different end-to-end deadlines for different types of input data items, and (ii) a new event model to represent the arrival pattern of the data items to be processed, which triggers the task structure. This event model is more expressive than classical models such as purely periodic, periodic with jitter or sporadic event models. We then present algorithms for the schedulability analysis of this task model. The basic scheme underlying our algorithms is a generalization of the techniques used for the schedulability analysis of the recently proposed generalized multiframe and the recurring real-time task models.

Multihomed network is one of the most efficient configuration to improve the response time of network services. However, it is hard to introduce or manage because the existing configuration methods have several problems in that they require much technical skill, involve administrative over-burden for the administrator and so on. In this paper, we propose a dynamic traffic sharing technique and suitable backbone selection metrics to address some of these problems. Using the proposed technique, an appropriate backbone can be selected per connection with minimal technical skill and low administrative cost. In addition, the proposed metrics performs more efficient traffic sharing as compared to others techniques that were investigated

The deployment of translucent optical networks is considered the most promising short term solution to decrease costs and energy consumption in optical backbone networks. Indeed, due to the technological maturity of translucent wavelength switched optical network (WSON) architectures, they have already caught close attention from the research community. Moreover, recent advances and enhancements in optical devices are now (re-)fostering research interest in sub-wavelength technologies like, among others, optical burst switching (OBS) and optical packet switching (OPS). Hence, in this paper, we evaluate and compare two novel node architectures for a translucent OBS(T-OBS) network. To be precise, we study nodes with both dedicated and shared wavelength converter resources (i.e. DWC and SWC). To this end, we consider the impact of the main physical layer impairments (PLIs) and make use of a routing and regenerator placement and dimensioning (RRPD)algorithm to minimize the number of optical-electrical-optical (O/E/O) regenerators deployed in the network whilst, at the same time, guaranteeing a target quality of transmission (QoT)network performance. The results presented prove the feasibility and the significant savings, in terms of the number of wavelength converters (WCs), that can be achieved by considering a network with shared wavelength converter resources.

Adaptive routing is an efficient congestion avoidance mechanism for modern Datacenter and HPC networks. Congestion detection traditionally relies on the occupancy of the router queues. However, this approach can hinder performance due to coarse-grain measurements with small buffers, and potential routing oscillations with large buffers. We introduce an alternative mechanism, labelled Contention-Based Adaptive Routing. Our mechanism adapts routing based on an estimation of “network contention”, the simultaneity of traffic flows contending for a network port. Our system employs a set of counters which track the demand for each output port. This exploits path diversity thanks to earlier detection of adversarial traffic patterns, and decouples buffer size and queue occupancy from contention detection. We evaluate our mechanism in a Dragonfly network. Our evaluations show this mechanism achieves optimal latency under uniform traffic and similar to best previous routing mechanisms under adversarial patterns, with immediate adaptation to traffic pattern changes. This work has been supported by the Spanish Ministry of Education, FPU grant FPU13/00337, the Spanish Science and Technology Commission (CICYT) under contracts TIN2012-34557 and TIN2013-46957-C2-2-P, the European Union FP7 under Agreement ICT-288777 (Mont-Blanc) and ERC-321253 (RoMoL), the European HiPEAC Network of Excellence, and the JSA no. 2013-119 as part of the IBM/BSC Technology Center for Supercomputing agreement.

Estimating the unknown parameters of a statistical model based on the observations collected by a sensor network is an important problem with application in multiple fields. In this setting, distributed processing, by which computations are carried out within the network in order to avoid raw data transmission to a fusion centre, is a desirable feature resulting in improved robustness and energy savings. In the presence of incomplete data, the expectation-maximisation (EM) algorithm is a popular means to iteratively compute the maximum likelihood (ML) estimate. It has found application in diverse fields such as computational biology, anomaly detection, speech segmentation, reinforcement learning, and motion estimation, among others. In this chapter we will review the formulation of the centralised EM estimation algorithm as a starting point and then discuss distributed versions well suited for implementation in sensor networks. The first class of these distributed versions requires specialised routing through the network in terms of a linear or circular path visiting all nodes, whereas the second class does away with this requirement by using the concept of network consensus to diffuse information through the network. Our focus will be on a relevant sensor network application, in which the parameter of a linear model is to be estimated in the presence of an unknown number of randomly malfunctioning sensors.

International audience; We consider the problem of finding a lightpath assignment for a given set of communication requests on a multifiber WDM optical network with wavelength translators. Given such a network and w, the number of wavelengths available on each fiber, k, the number of fibers per link, and c, the number of partial wavelength translations available on each node, our problem stands for deciding whether it is possible to find a w-lightpath for each request in the set such that there is no link carrying more that k lightpaths using the same wavelength nor node where more than c wavelength translations take place. Our main theoretical result is the writing of this problem as a particular instance of integral multicommodity flow, hence integrating routing and wavelength assignment in the same model. We then provide three heuristics mainly based upon randomized rounding of fractional multicommodity flow and enhancements that are three different answers to the trade-off between efficiency and tightness of approximation, and discuss their practical performances on both theoretical and real-world instances.

Using forward error correction (FEC) in off-line streaming, permitting large buffers, yields spectacular results, but real-time streaming puts hard restrictions on the buffer size and therefore does not allow FEC to deal with long link failures on a single path route. Multi-path routing however can make FEC efficient also for realtime streaming. In this paper, we introduce a capillary routing algorithm offering a wide range of multi-path routing topologies starting from a simple (max-flow multi-path) solution toward more reliable and secure schemes obtained by recursively spreading individual sub-flows. The friendliness of a particular multi-path routing can be measured by adaptive redundancy overall need (ARON), which is proportional to the sender's total channel coding effort needed for recovering the failure of each individual link in the multi-path route. A dozen of capillary routing layers are built on several hundreds of network samples obtained from a random walk wireless mobile ad-hoc network (MANET). Rating of these routing suggestions with ARON shows that the FEC friendliness improves substantially as the spreading of the routing grows

In off-line streaming, packet level erasure resilient forward error correction (FEC) codes rely on the unrestricted buffering time at the receiver. In real-time streaming, the extremely short playback buffering time makes FEC inefficient for protecting a single path communication against long link failures. It has been shown that one alternative path added to a single path route makes packet level FEC applicable even when the buffering time is limited. Further path diversity, however, increases the number of underlying links increasing the total link failure rate, requiring from the sender possibly more FEC packets. We introduce a scalar coefficient for rating a multi-path routing topology of any complexity. It is called redundancy overall requirement (ROR) and is proportional to the total number of adaptive FEC packets required for protection of the communication. With the capillary routing algorithm, introduced in this paper we build thousands of multi-path routing patterns. By computing their ROR coefficients, we show that contrary to the expectations the overall requirement in FEC codes is reduced when the further diversity of dual-path routing is achieved by the capillary routing algorithm

We show how inaccuracies of Quality of Transmission (QoT) estimations, caused by imperfect models and lack of monitors in transparent optical networks, can be mitigated using a novel routing and wavelength assignment algorithm.

We address the design of a packet buffer for future high-speed routers that support line rates as high as OC-3072 (160 Gb/s), and a high number of ports and service classes. We describe a general design for hybrid DRAM/SRAM packet buffers that exploits the bank organization of DRAM. This general scheme includes some designs previously proposed as particular cases. Based on this general scheme, we propose a new scheme that randomly chooses a DRAM memory bank for every transfer between SRAM and DRAM. The numerical results show that this scheme would require an SRAM size almost an order of magnitude lower than previously proposed schemes without the problem of memory fragmentation.