Search

Mobile communications are a powerful contributor to social and economic develop- ment worldwide, including in less developed or remote parts of the world. However they are large users of electricity through their base stations, backhaul networks and Cloud servers, so that they have a large environmental impact when they use the electric grid. On the other hand, they could operate with renewable energy sources and thus reduce their CO2 impact and be accessible even in areas where the electric grid is unavailable or unreliable. The counterpart is that intermittent sources of energy, such as photovoltaic and wind, can affect the quality of service (QoS) that is experienced by mobile users. Thus in this paper we model the per- formance of mobile telecommunications that use intermittent and renewable energy sources. In such cases to analyse the performance of such systems, both the energy supply and the network traffic, can be modeled as random processes, and we develop mathematical models using the Energy Packet Network paradigm, where both data and energy flows are discretised. QoS metrics for the users are computed based on the traffic intensity and the availability of energy. &nbsp

The Cognitive Packet Network (CPN) bases its routing decisions and flow control on the Random Neural Network (RNN) Reinforcement Learning algorithm; this paper proposes the addition of a Deep Learning (DL) Cluster management structure to the CPN for Quality of Service metrics (Delay Loss and Bandwidth), Cyber Security keys (User, Packet and Node) and Management decisions (QoS, Cyber and CEO). The RNN already models how neurons transmit information using positive and negative impulsive signals whereas the proposed additional Deep Learning structure emulates the way the brain learns and takes decisions; this paper presents a brain model as the combination of both learning algorithms, RNN and DL. The proposed model has been simulated under different network sizes and scenarios and it has been validated against the CPN itself without DL clusters. The simulation results are promising; the presented CPN with DL clusters as a mechanism to transmit, learn and make packet routing decisions is a step closer to emulate the way the brain transmits information, learns the environment and takes decisions.

We present an approach that uses Reinforcement Learning (RL) with the Random Neural Network (RNN) acting as an adaptive critic, to route traffic in a SDN network, so as to minimize a composite Goal function that includes both packet delay and energy consumption per packet. We directly measure the traffic dependent energy consumption characeristics of the hardware that we use (including energy expended per packet) so as to parametrize the Goal function. The RL based algorithm with the RNN is implemented in a SDN controller that manages a multi-hop network which assigns service requests to specific servers so as to minimize the desired Goal. The overall system’s performance is evaluated through experimental measurements of packet delay and energy consumption under different traffic load values, demonstrating the effectiveness of the proposed approach.

Though currently a “hot topic”, over the past fifteen years [1][2], there has been significant work on the use of machine learning to design large scale computer-communication networks, motivated by the complexity of the systems that are being considered and the unpredictability of their workloads. A topic of great concern has been security [3] and novel techniques for detecting network attacks have been developed based on Machine Learning [8]. However the main challenge with Machine Learning methods in networks has concerned their compatibility with the Internet Protocol and with legacy systems, and a major step forward has come from the establishment of Software Defined Networks (SDN) [4] which delegate network routing to specific SDN routers [4]. SDN has become an industry standard for concentrating network management and routing decisions within specific SDN routers that download the selected paths periodically to network routers, which operate otherwise under the IP protocol. In this paper we describe our work on real-time control of Security and Privacy [7], Energy Consumption and QoS [6] of packet networks using Machine Learning based on the Cognitive Packet Network [9] principles and their application to the H2020 SerIoT Project [5].

Using the Energy Packet Network (EPN) model, we show how energy can be shared between heterogeneous servers at the edge to minimize the overall average response time of jobs. The system is modelled as a probabilistic network where energy and jobs are being dispatched to the edge servers using G-Networks with a product-form solution for the equilibrium probability distribution of system state. The approach can also be used to design energy dispatching systems when renewable energy is used to improve the sustainability of edge computing.

Over the last decade, the increased automation in the management of computer systems and networks, the rising costs of energy, and the sensitivity about the environmental impact of energy consumption, has resulted in an increase of the importance of energy in the overall running of ICT (Information and Communication) Systems. The deregulation in telecommunication services has also had an impact in this direction because of the increased duplication in telecommunication infrastructures. Furthermore, most cyberattacks, also increase the energy consumption of ICT systems. Thus, today the overall electricity consumed by ICT worldwide is comparable to the total electricity consumption of two major industrial powers: Japan plus Germany. Clearly, of one wishes to save energy in ICT, this may come at the expense of a reduction in the measured Quality of Service (QoS) of ICT systems. Thus we will survey a number of our own research results, covering wireless communications, compute servers and network routing, to offer optimal trade-offs between energy consumption and QoS. The Energy Packet Network (EPN) that we have introduced to optimise the Energy-QoS trade-offs will be described and applied to a number of examples. We will also describe some results on the potential and limitations for communicating and computing with particle spins as a means to achieve energy savings.

We investigate how the flow of energy and the flow of jobs in a service system can be used to minimize the average response time to jobs that arrive according to random arrival processes at the servers. An interconnected system of workstations and energy storage units that are fed with randomly arriving harvested energy is analyzed by means of the Energy Packet Network (EPN) model. The system state is discretized, and uses discrete units to represent the backlog of jobs at the workstations, and the amount of energy that is available at the energy storage units. An Energy Packet (EP) which is the unit of energy, can be used to process one or more jobs at a workstation, and an EP can also be expended to move a job from one workstation to another one. The system is modeled as a probabilistic network that has a product-form solution for the equilibrium probability distribution of system state. The EPN model is used to solve two problems related to using the flow of energy and jobs in a multi-server system, so as to minimize the average response time experienced by the jobs that arrive at the system.

International audience; In this paper, we analyze the network attacks that can be launched against Internet of Things (IoT) gateways, identify the relevant metrics to detect them, and explain how they can be computed from packet captures. We then present the principles and design of a deep learning-based approach using dense random neural networks (RNN) for the online detection of network attacks. Empirical validation results on packet captures in which attacks are inserted show that the Dense RNN correctly detects attacks. However our experiments show that a simple threshold detector also provides results of comparable accuracy on the same data set.

In this paper, we analyze the network attacks that can be launched against IoT gateways, identify the relevant metrics to detect them, and explain how they can be computed from packet captures. We also present the principles and design of a deep learning-based approach using dense random neural networks (RNN) for the online detection of network attacks. Empirical validation results on packet captures in which attacks were inserted show that the Dense RNN correctly detects attacks.

We consider a wireless sensor node that gathers energy through harvesting and reaps data through sensing. The node has a wireless transmitter that sends out a data packet whenever there is at least one “energy packet” and one “data packet”, where an energy packet represents the amount of accumulated energy at the node that can allow the transmission of a data packet. We show that such a system is unstable when both the energy storage space and the data backlog buffer approach infinity, and we obtain the stable stationary solution when both buffers are finite. We then show that if a single energy packet is not sufficient to transmit a data packet, there are conditions under which the system is stable, and we provide the explicit expression for the joint probability distribution of the number of energy and data packets in the system. Since the two flows of energy and data can be viewed as flows that are instantaneously synchronised, this paper also provides a mathematical analysis of a fundamental problem in computer science related to the stability of the “join” synchronisation primitive.

This paper develops multi-layer classifiers and auto-encoders based on the Random Neural Network. Our motivation is to build robust classifiers that can be used in systems applications such as Cloud management for the accurate detection of states that can lead to failures. Using an idea concerning some to soma interactions between natural neuronal cells, we discuss a basic building block constructed of clusters of densely packet cells whose mathematical properties are based on G-Networks and the Random Neural Network. These mathematical properties lead to a transfer function that can be exploited for large arrays of cells. Based on this mathematical structure we build multi-layer networks. In order to evaluate the level of classification accuracy that can be achieved, we test these auto-encoders and classifiers on a widely used standard database of handwritten characters.

This paper develops a mathematical model to determine the balance of energy input and data sensing and transmission in a wireless sensing node. Since the node acquires energy through harvesting from an intermittent source, and sensing is also carried out intermittently, the node is modelled with random arrivals of both energy and data. A buffer in the node stores data packets while energy is stored in a battery acting as an energy buffer. The approach uses the “Energy Packet Network” paradigm so that both energy and data packets can be modelled as discrete quantities. We assume that for each data packet, the sensor consumes K e energy packets for node electronics including sensing, processing, and storing and K t energy packets for transmission. We model the node's energy and data flow by a two-dimensional random walk which represents the backlog of data and energy packets. We then simplify the model using companion matrices and matrix algebra techniques that allow us to obtain a closed-form solution for the stationary probability distribution for the random walk which allows us to compute important performance measures, including the energy consumed by the node, and its throughput in data packets transmitted as a function of the amount of power that it receives. The model also allows us to evaluate the effect of ambient noise and the needs for data retransmissions, including for the case where M sensors operate in proximity and create interference for each other.

Energy saving in networks has traditionally focussed on reducing battery consumption through smart wireless network design. Recently, researchers have turned their attention to the energy cost and carbon emissions of the backbone network that both fixed and mobile communications depend on, proposing primarily mechanisms that turn equipments OFF or put them into deep sleep. This is an effective way of saving energy, provided that the nodes can return to working condition quickly, but it introduces increased delays and packet losses that directly affect the quality of communication experienced by the users. Here we investigate the associated tradeoffs between power consumption and quality of service in backbone networks that employ deep sleep energy savings. We examine these tradeoffs by conducting experiments on a real PC-based network topology, where nodes are put into deep sleep at random times and intervals, resulting in a continuously changing network with reduced total power consumption. The average power consumption, the packet loss and the average delay of this network are examined with respect to the average value of the ON rate and the ON/OFF cycle of the nodes.

This paper discusses a two-dimensional random walk for modeling a data transmission system with energy harvesting that represents a remotely operating wireless sensor node. The node has a wireless transmitter that gathers data from the environment according to a random process, and similarly it harvests energy from the environment at a random rate. Then if it has gathered enough data, it transmits a data packet provided it has stored at least K ≥ 1 energy packets that are needed to transmit one data packet. Generalising previous results that were obtained for the case K = 1, we derive the stationary probability distribution of this model to compute principal performance metrics for the wireless sensor node with energy harvesting, including the average transmission power emanating from any sensor node. We then compute the probability that a bit is correctly received by a receiver that operates in the presence of N identical wireless sensors, each operating at the power level K, in the presence of noise and of interference due to the transmissions. Numerical results show how the noise power B and the transmission power K affect the probability of correctly receiving a bit in such a system.

The flow of energy in a network of intermittent sources of energy, such as renewables, batteries, and of work such as sensing tasks, communication packets and tasks executed in computer servers, is modeled as an Energy Packet Network. Both tasks for execution and energy flows into the system are random processes. A G-Network is used to model the system where both tasks and data packets, and energy in the form of discrete units, move between energy storage units, network routers and computing servers. This queueing network approach offers the advantage of leading to an analytical solution of the equilibrium joint probability distribution for the backlog of work and the amount of energy that is stored in the system. We then consider the problem of deciding how energy and workload should be allocated among the workstations in the system so as to optimise utility functions that combine the need to offer short response times to computing jobs and data packets, while maintaining some level of reserve energy in storage for potential future needs. We propose a gradient based approach to exploit the continuous and differentiable structure of the utility function and suggest an approximate gradient algorithm based on a Neumann expansion to simplify the computations. The performance of the approximate algorithms is compared numerically to the case where an exact gradient algorithm is used, both for the computational cost and the resulting minimum value of the utility function.

© Springer International Publishing Switzerland 2016.Real-Time services over IP (RTIP) have been increasingly significant due to the convergence of data networks worldwide around the IP standard, and the popularisation of the Internet. Real-Time applications have strict Quality of Service (QoS) constraint, which poses a major challenge to IP networks. The Cognitive Packet Network (CPN) has been designed as a QoS-driven protocol that addresses user-oriented QoS demands by adaptively routing packets based on online sensing and measurement, and in this paper we design and experimentally evaluate the “Real-Time (RT) over CPN” protocol which uses QoS goals that match the needs of real-time packet delivery in the presence of other background traffic under varied traffic conditions. The resulting design is evaluated via measurements of packet delay, delay variation (jitter) and packet loss ratio.

This paper examines the use of a gradient-based algorithm for Quality of Service (QoS) and power minimisation in wired networks to result on reduced energy consumption. Two distinct schemes, conventional shortest-path routing and an autonomic algorithm energy aware routing algorithm (EARP) are investigated as the starting point for the gradient algorithm. Comparisons are conducted using the same network test-bed and identical network traffic under conditions where routers and link drivers are always kept on so as to meet the needs for network reliability in the presence of possible failures and unexpected overload. Since splitting traffic flows can increase jitter and the arrival of packets which are out of sequence, we also do not allow the same packet flow to be conveyed over multiple paths. In the experiments that we have conducted we observe that power consumed with the gradient-optimiser that is proposed in this paper is a few percent to 10% smaller than that consumed using shortest-path routing or EARP. Although the savings are small, they can be very significant for the Internet backbone as a whole over long periods of time, and further power savings may be obtained by judiciously putting to sleep equipment when it is under-utilised.

Introduction Applications including remote sensing, embedded sensing (e.g. in bridges or buildings), or sensors for public safety require that sensing and transmission take place with minimal human intervention and no direct connection to the power mains. Ideally, sensing nodes should be autonomous and able to harvest ambient energy from vibrations, light, electromagnetic radiation, or heat. Thus energy harvesting nodes for wireless communications [5, 1] are of great interest. Here we study nodes that have packet buffers and energy storage buffers (e.g. capacitors or batteries) and we use the Energy Packet(EP) [2, 3] abstraction so that one EP is the amount of energy needed to transmit a Data Packet (DP). Ideally the energy flow into the device, expressed as EPs per unit time, should balance the flow of DPs that are assembled for transmission so that DPs are not held up. However we show that if EP and DP flows are exactly balanced, the system exhibits an unstable behaviour so that both the backlog of packets and the amount of stored energy saturate their respective storage, and derive results concerning the performance of the system when the energy and work flows are unbalanced. The study can be generalised to the multihop case. Consider a system that at stime t contains K(t) DPs to transmit and M(t) EPs in storage. Let the data buffer capacity be B DPs, while the energy storage device can store up to E EPs. The state of the system is then the pair (K(t),M(t)) and are interested in its probability distribution p(n,m, t) = Pr[K(t) = n,M(t) = m). Arrival of DPs to the data buffer follows a Poisson process of rate λ, while the EPs arrive into the energy buffer as a Poisson process of rate Λ corresponding to some rate of a joules/second, where a is the energy required to transmit one packet. We assume that the rates at which energy is harvested and data collected is slow as compared to the rate at which DPs are transmitted by the device, e.g. it may take many milliseconds to get enough energy from the harvester while it may

Although there is great interest in reducing energy consumption for all areas of human activity, many of the proposed approaches such as the smart home or the smart grid are actually prone to an increase in the use of Information and Communication Technologies (ICT) which itself is a big consumer of energy. It is remarkable that ICT's carbon imprint is of the order of 2% of the world total, comparable to the carbon imprint of air travel. Thus, it is imperative to address energy savings in ICT and, in particular, in data centres and networks. This paper follows up on our previous work that seeks novel ways to reduce the energy consumption in packet networks which constitute the backbone of the Internet and of the information society as a whole. Here we discuss the use of routing control as a means to reduce energy consumption while remaining aware of QoS considerations, and propose a method that uses a queueing theoretic analysis and optimization technique to distribute traffic so as to reduce a cost function that comprises both energy and QoS. A method using G-networks is developed to incorporate both the effect of user traffic and the overhead in QoS and energy consumption introduced by the control traffic that will be needed to carry out the re-routing decisions. For an N-node network, we show that this approach results in an algorithm which has O(N3) time complexity. Because this approach may be too costly in computational overhead and delays, we also propose another approach that uses load balancing and which would be much simpler to implement.

Large wireless sensor networks consisting of hundreds of devices are often used to conduct distinct monitoring tasks simultaneously. Furthermore, the significant events that can be detected by the network may be detected in many different parts of the network. As a consequence, these multiple tasks and diverse events may also cause resource contention and degradation of network quality of service. For instance, congestion created by regularly occurring events which do not need urgent forwarding to the monitoring centres may create delays for other events which need to be reported very rapidly. The Random Re-Routing (RRR) protocol for wireless sensor networks has therefore been proposed to adaptively prioritise the packets of certain urgent events above those of other less urgent events which may be of a more routine nature. In this paper we extend the RRR protocol to improve performance, and we use simulation experiments to illustrate heuristic algorithms for matching routing priorities to the requirements of the sensor network's monitoring tasks. The approaches are based on selecting the protocol's parameters prior to network deployment, and on using an acknowledgement mechanism for adapting the parameters during network operation.

New approaches to Quality-of-Service (QoS) routing in wireless sensor networks which use different forms of learning are the subject of this paper. The Cognitive Packet Network (CPN) algorithm uses smart packets for path discovery, together with reinforcement learning and neural networks, while Self-Selective Routing (SSR) is based on the “Ant Colony” paradigm which emulates the pheromone-based technique which ants use to mark paths and communicate information about paths between different insects of the same colony (Koenig et al. in Ann Math Artif Intell 31(1–4): 41–76, 2001). In this paper, we present first experimental results on a network test-bed to evaluate CPN’s ability to discover paths having the shortest delay, or shortest length. Then, we present small test-bed experiments and large-scale network simulations to evaluate the effectiveness of the SSR algorithm. Finally, the two approaches are compared with respect to their ability to adapt as network conditions change over time.

Smart networks have grown out of the need for stable, reliable, and predictable networks that will guarantee packet delivery under Quality of Service (QoS) constraints. In this article we present a measurement-based admission control algorithm that helps control traffic congestion and guarantee QoS throughout the lifetime of a connection. When a new user requests to enter the network, probe packets are sent from the source to the destination to estimate the impact that the new connection will have on the QoS of both the new and the existing users. The algorithm uses a novel algebra of QoS metrics, inspired by Warshall's algorithm, to look for a path with acceptable QoS values to accommodate the new flow. We describe the underlying mathematical principles and present experimental results obtained by evaluating the method in a large laboratory test-bed operating the Cognitive Packet Network (CPN) protocol.

Denial of service (DoS) attacks are a serious security threat for Internet based organisations, and effective methods are needed to detect an attack and defend the nodes being attacked in real time. We propose an autonomic approach to DoS defence based on detecting DoS flows, and adaptively dropping attacking packets upstream from the node being attacked using trace-back of the attacking flows. Our approach is based on the Cognitive Packet Network infrastructure which uses smart packets to select paths based on Quality of Service. This approach allows paths being used by a flow (including an attacking flow) to be identified, and also helps legitimate flows to find robust paths during an attack. We evaluate the proposed approach using a mathematical model, as well as using experiments in a laboratory test-bed. We then suggest a more sophisticated defence framework based on authenticity tests as part of the detection mechanism, and on assigning priorities to incoming traffic and rate-limiting it on the basis of the outcome of these tests.

We present a performance model for an energy harvesting wireless sensor node in which data gathering and harvesting are slow random processes as compared to fast wireless communications. We assume that the system will use stored energy when collecting data in standby, and that energy will leak from capacitors and batteries. In the presence of these imperfections we derive the system's packet transmission capacity when its packet storage buffer and its energy storage unit have a finite capacity that may lead to both data packet overflows, and the loss of incoming energy in addition to standby losses. We also consider an infinite capacity model which operates in the presence of transmission errors due to channel noise and interference.

We demonstrate experimentally how an Autonomic Network based on the CPN protocol can provide the Quality of Service (QoS) required by voice communications. The implementation uses Reinforcement Learning to dynamically seek paths that meet the quality requirements of voice communications. Measurements of packet delay, jitter, and loss illustrate the performance obtained from the system.

We consider an interconnected distributed computer system with multiple computation centres (CC) that operate with energy harvesting to improve sustainability. The intermittent energy harvesting is matched with steady demand from the CCs using energy storage (ES), e.g. batteries. Based on energy leakage from batteries, and power losses over transmission lines, we examine whether a centralised or distributed ES system provides the solution that offers the smallest response time to a fixed workload of computer jobs using the Energy Packet Network (EPN) modelling paradigm.

This paper studies interconnected wireless sensors with the paradigm of Energy Packet Networks (EPN) which were previously introduced. In the EPN model, both data transmissions and the flow of energy are discretized, so that an energy packet (EP) is the minimum amount of energy (say in microjules) that is needed to process and transmit a data packet (DP) or to process a job. Previous work has modeled such systems to determine the relation between energy flow and DP transmission, or to study the balance between energy and the processing of jobs in Cloud Servers. The lack of energy, in addition to processing times, is the main source of latency in networks of sensor nodes. Thus this paper models this phenomenon, and shows that under some reasonable conditions, assuming feedforward flow of data packets and local consumption and leakage of energy, such networks have product form solutions.

Packet routing in networks requires knowledge about available paths, which can be either acquired dynamically while the traffic is being forwarded, or statically (in advance) based on prior information of a network's topology. This paper describes an experimental investigation of path discovery using genetic algorithms (GAs). We start with the quality-of-service (QoS)-driven routing protocol called "cognitive packet network" (CPN), which uses smart packets (SPs) to dynamically select routes in a distributed autonomic manner based on a user's QoS requirements. We extend it by introducing a GA at the source routers, which modifies and filters the paths discovered by the CPN. The GA can combine the paths that were previously discovered to create new untested but valid source-to-destination paths, which are then selected on the basis of their "fitness." We present an implementation of this approach, where the GA runs in background mode so as not to overload the ingress routers. Measurements conducted on a network test bed indicate that when the background-traffic load of the network is light to medium, the GA can result in improved QoS. When the background-traffic load is high, it appears that the use of the GA may be detrimental to the QoS experienced by users as compared to CPN routing because the GA uses less timely state information in its decision making.

We present a novel direction based shortest path search algorithm to guide evacuees during an emergency. It uses opportunistic communications (oppcomms) with low-cost wearable mobile nodes that can exchange packets at close range of a few to some tens of meters without help of an infrastructure. The algorithm seeks the shortest path to exits which are safest with regard to a hazard, and is integrated into an autonomous Emergency Support System (ESS) to guide evacuees in a built environment. The ESS that we propose, that includes the directional algorithm and the Oppcomms, are evaluated using simulation experiments with the DBES (Distributed Building Evacuation Simulator) tool by simulating a shopping centre where fire is spreading. The results show that the directional path finding algorithm can offer significant improvements for the evacuees. In particular, we see that the improved and more reliable communications offered by Oppcomms, especially when the number of evacuees is larger, can help to compensate for the effects of congestion and improve the overall success of the evacuation scheme. Throughout the simulations we observe improvements of a few percent, which can translate into a valuable number of more people that are safely evacuated when human lives and safety are at risk.

This paper proposes a new energy efficient algorithm to find and maintain routes in mobile ad hoc networks. The proposal borrows the notion of learning from a previous research on cognitive packet networks (CPN) to create a robust routing protocol. Our idea uses smart packets that exploit the use of unicasts and broadcasts to search for routes. Because unicasts impose lower overall overhead, their use is preferred. Smart packets learn how to make good unicast routing decisions by employing a combined goal function which considers both the energy stored in the nodes and path delay. The end result is a dynamic discovery of paths that offer an equilibrium between low-delay routes and an efficient use of network resources that extends the working lifetime of the network.

The use of wireless sensor networks (WSNs) for emergency navigation systems suffer disadvantages such as limited computing capacity, restricted battery power and high likelihood of malfunction due to the harsh physical environment. By making use of the powerful sensing ability of smart phones, this paper presents a cloud-enabled emergency navigation framework to guide evacuees in a coordinated manner and improve the reliability and resilience in both communication and localization. By using social potential fields (SPF), evacuees form clusters during an evacuation process and are directed to egresses with the aid of a Cognitive Packet Networks (CPN) based algorithm. Rather than just rely on the conventional telecommunications infrastructures, we suggest an Ad hoc Cognitive Packet Network (AHCPN) based protocol to prolong the life time of smart phones, that adaptively searches optimal communication routes between portable devices and the egress node that provides access to a cloud server with respect to the remaining battery power of smart phones and the time latency., Comment: This document contains 8 pages and 3 figures and has been accepted by ISCIS 2014 (29th International Symposium on Computer and Information Sciences)

We present a novel direction based shortest path search algorithm to guide evacuees during an emergency. It uses opportunistic communications (oppcomms) with low-cost wearable mobile nodes that can exchange packets at close range of a few to some tens of meters without help of an infrastructure. The algorithm seeks the shortest path to exits which are safest with regard to a hazard, and is integrated into an autonomous Emergency Support System (ESS) to guide evacuees in a built environment. The algorithm proposed that ESSs are evaluated with the DBES (Distributed Building Evacuation Simulator) by simulating a shopping centre where fire is spreading. The results show that the directional path finding algorithm can offer significant improvements for the evacuees.

This paper explores the idea of smart building evacuation when evacuees can belong to different categories with respect to their ability to move and their health conditions. This leads to new algorithms that use the Cognitive Packet Network concept to tailor different quality of service needs to different evacuees. These ideas are implemented in a simulated environment and evaluated with regard to their effectiveness., 7 pages, 7 figures

Voice over IP (VOIP) has strict Quality of Service (QoS) constraints and requires real-time packet delivery, which poses a major challenge to IP networks. The Cognitive Packet Network (CPN) has been designed as a QoS-driven protocol that addresses user-oriented QoS demands by adaptively routing packets based on online sensing and measurement. This paper presents our design, implementation, and evaluation of a ‘Voice over CPN’ system where an extension of the CPN routing algorithm has been developed to support the needs of voice packet delivery in the presence of other background traffic flows with the same or different QoS requirements.

As the size of the Internet grows by orders of magnitude both in terms of users, number of IP addresses, and number of routers, and as the links we use (be they wired, optical or wireless) continuously evolve and provide varying reliability and quality of service, the IP based network architecture that we know so well will have to evolve and change. Both scalability and QoS have become key issues. We are currently conducting a research project that revisits the IP routing architecture issues and proposes new designs for routers. As part of this effort, this paper discusses a packet network architecture called a cognitive packet network (CPN), in which intelligent capabilities for routing and flow control are moved towards the packets, rather than being concentrated in the nodes. In this paper we outline the design of the CPN architecture, and discuss the quality-of-service based routing algorithm that we have designed and implemented. We then present our test-bed and report on extensive measurement experiments that we have conducted.

We discuss a packet network architecture called a cognitive packet network (CPN), in which intelligent capabilities for routing and flow control are moved towards the packets, rather than being concentrated in the nodes and protocols. Our architecture contains “smart” and “dumb” packets, as well as acknowledgement packets. Smart CPN packets route themselves, and learn to avoid congestion and losses from their own observations about the network and from the experience of other packets. They use a reinforcement learning algorithm to route themselves based on a goal function which has been assigned to them for each connection. Dumb CPN packets of a specific quality of service (QoS) class use routes which have been selected by the smart packets (SPs) of that class. Acknowledgement (ACK) packets are generated by the destination when an SP arrives there; the ACK heads back to the source of the SP along the inverse route and is used to update mailboxes in CPN routers, as well as to provide source routing information for dumb packets. We first summarize the basic concepts behind CPN, and present simulations illustrating their performance for different QoS goals, and analytical results for best and worst case performance. We then describe a test-bed network we have designed and implemented in order to demonstrate these ideas. We provide measurement data on the test-bed to illustrate the capacity of the network to adapt to changes in traffic load and to failures of links. Finally, we use measurements to evaluate the impact of the ratio of smart to dumb packets on the end-to-end delay experienced by all of the packets.

A cognitive user (CU) operates as a secondary user of a cognitive channel. Before transmission CU samples the channel until it estimates that the channel can be accessed successfully. When CU transmits a packet, it may nevertheless be unsuccessful because its estimate was wrong; CU knows about its failure and stops the ongoing transmission sometime before the transmission ends. Then CU restarts sensing the channel. In this paper we derive the first and second moments of the effective transmission time of a packet which experiences additional delay which is similar to a “vacation time” in queues, though it differs from conventional vacation time models because such “sampling vacations” are beneficial to the future transmission. Using an on-off model of channel occupancy, our analysis shows that the duration of the sensing and the packet's length can both be tuned to optimize for throughput, energy, interference, quality of service (QoS) and average response time of CUs.

We consider a smart or cognitive user (CU) that operates as a secondary user of a cognitive channel. Before transmission, the CU samples the channel until it estimates that it can be accessed successfully. When the CU transmits a packet, it may nevertheless be unsuccessful because its estimate was wrong. The CU then knows about its failure and stops the ongoing transmission sometime before the transmission ends. Then the CU restarts sensing the channel. In this paper we analyse the total delay experienced by a CU. We derive the first and second moments of the effective transmission time of a packet sent by such a smart user, in view of the fact that the sensing and errors made in transmitting a packet when the channel is actually unavailable, will introduce additional delay. This is similar to a “vacation time” in queues, though it differs from conventional vacation time models because such “sampling vacations” can be beneficial to the future transmission.

Providing optimal and safe routes to evacuees in emergency situations requires fast and adaptive algorithms. The common approaches are often too slow to converge, too complex, or only focus on one aspect of the problem, e.g. finding the shortest path. This paper presents an adaptation of the Cognitive Packet Network (CPN) concept to emergency evacuation problems. Using Neural Networks, CPN is able to rapidly explore a network and allocate overhead in proportion to the perceived likelihood of finding an optimal path there. CPN is also flexible, as it can operate with any user-defined cost function, such as congestion, path length, safety, or even compound metrics. We compare CPN with optimal algorithms such as Dijkstra’s Shortest Path using a discrete-event emergency evacuation simulator. Our experiments show that CPN reaches the performance of optimal path-finding algorithms. The resulting side-effect of such smart or optimal algorithms is in the greater congestion that is encountered along the safer paths; therefore we indicate how the quality of service objective used by CPN can also be used to avoid congestion for further improvements in evacuee exit times.

Network testbeds are useful for protocol performance evaluation. They overcome most challenges commonly involved in live network experimentation, retaining fair realism and repeatability under controllable conditions. However, large-scale testbeds are difficult to set up due to limited resources available to most experimenters. In this paper, we explore the use of hardware virtualisation as an experimental tool to improve resource efficiency, allowing to boost the effective number of nodes available for network testing. By virtualising network routers and links, a cluster environment with off-the-shelf equipment can host hundreds of virtual routers for large-scale network testing. We apply this technique to construct a 800-node Cognitive Packet Networks (CPN) testbed that provides insight into the benefits and limitations of the approach.

We describe an autonomous emergency support system (ESS) based on opportunistic communications (oppcomms) to support navigation and evacuation of civilians in built environments. In our proposed system, civilians are equipped with low-cost human wearable devices that employ oppcomms to exchange packets at close range of a few meters with limited or no infrastructure. This paper investigates the resilience and performance of oppcomms in the presence of network attacks. We assume that a portion of nodes in the network have been compromised and misbehave, which adversely affects communications and evacuation. We evaluate the effect of three types of node misbehaviour and propose a defense mechanism against the most serious among these. The defense mechanism combines identity-based cryptography with collaborative malicious packet detection and blacklisting of detected attackers. Results from simulation experiments conducted on a specialized emergency simulator show the impact of misbehaviour on evacuation and communication performance and the improvement offered by the defense mechanism.

In the presence of limitations in the availability of energy for data centres, especially in dense urban areas, a novel system that we call an Energy Packet Network is discussed as a means to provide energy on demand to Cloud Computing servers. This approach can be useful in the presence of renewable energy sources, and if scarce sources of energy must be shared by multiple computational units whose peak to average power consumption ratio is high. Such a system will use energy storage units to best match and smooth the intermittent supply and the intermittent demand. The analysis of such systems based on queueing networks is suggested and applied to a special case for illustration.

In the presence of limitations in the availability of energy for data centres, especially in dense urban areas, a novel system that we call an Energy Packet Network is discussed as a means to provide energy on demand to Cloud Computing servers. This approach can be useful in the presence of renewable energy sources, and if scarce sources of energy must be shared by multiple computational units whose peak to average power consumption ratio is high. Such a system will use energy storage units to best match and smooth the intermittent supply and the intermittent demand. The analysis of such systems based on queueing networks is suggested and applied to a special case for illustration.

The purpose of this paper is to examine how a gradient-based algorithm that minimises a cost function that includes both quality of service (QoS) and power minimisation in wired networks can be used to improve energy savings with respect to shortest-path routing, as well as against a "smart" autonomic algorithm called EARP which uses adaptive reinforcement learning. Comparisons are conducted based on the same test-bed and identical network trac. We assume that due to the need for network reliability and resilience we are not allowed to turn o routers and link drivers. We also assume that for QoS reasons (notably with regard to jitter and to avoid packet desequencing) we are not al- lowed to split trac from the same ow into dierent paths. Under these assumptions and for the considered trac, we observe that power con- sumed with the gradient-optimiser is a few percent to 10% smaller than that consumed using shortest-path routing or EARP. Since the magni- tude of the savings is small, this suggests that further power savings may only be obtained if under-utilised equipment can be dynamically put to sleep or turned on.

As packet networks become extremely large, as their applications and end users become more diverse and extremely numerous, a design based on a single view of how a network should operate becomes harder to justify. Thus each part of the network should be aware of its environment, of its users, of its resources, of its performance, of the energy it is consuming and of the threats that it is facing, and should use this awareness as an operational means to adapt its behaviour accordingly. This tutorial paper sets the stage for these ideas, and shows a simple and practical way forward based one the Cognitive Packet Network routing algorithm and the notion of Self-Aware Networks.

We formulate the problem of joint energy and quality ofservice (QoS) optimisation in packet networks and show how it can beformalised using the control capabilities inherent in G-network theory,which also includes the overhead due to control traffic. Using known energy consumption characteristics of network routers and of network linkdriver equipment, this approach leads to a polynomial time complexitygradient optimisation algorithm which seeks a judicious balance betweenthe QoS experienced by the users and the overall energy consumption ofthe network.

Admission control based on multiple criteria has always been a desirable feature for computer networks. In the past, we presented such a system, but it was based on a centralised decision mechanism, which by itself constitutes a significant limitation. Here, we describe a decentralised version of the initial admission control algorithm, while still keeping the multiple-criteria aspect, with each user being able to specify their Quality of Service (QoS) metrics at the level of the individual. Our scheme decides whether a new call should be allowed to enter the network based on measurements of the QoS metrics on each link of the network before and after the transmission of probe packets. The decision is based on a novel algebra of QoS metrics, inspired by Warshall's algorithm that searches whether there is a feasible path to accommodate the new flow without affecting the existing users. The decision is made at each source node individually, based on either only personal information or also on information exchange among the nodes that are involved. The performance of the algorithm is evaluated in terms of QoS throughout the lifetime of all connections. The experiments presented in this paper were conducted in an actual laboratory test-bed of realistic topology and under highly congested circumstances. The results are particularly encouraging.

We consider the problem of optimizing the perfor- mance of a network coding router with two stochastic flows. We develop a queueing model which accounts for the fact that coding is not performed when packets are transmitted, but is done by a separate program or hardware which operates independently of the hardware that sends packets out over links. We formulate and solve a constrained optimization problem which provides the optimal time that the router should wait before sending the information that it has uncoded, so that the average response time of the system is minimized. The trade-offs between delay and bandwidth or energy associated with the choice of the waiting time are also investigated, and the results indicate that network coding offers significant performance gains in a moderate to heavily loaded system.

Environmental monitoring is one of the most popular applications in wireless sensor networks. Although it is important to obtain a continuous record of the environment, users may gain enough information without receiving every routine sensor measurement. Reducing unnecessary traffic allows better utilization of the network resources. However, it is uneasy to decide on when and what kind of traffic to reduce in a dynamically changing environment. In this paper, we study the problem of information-aware traffic reduction for wireless sensor networks. We propose a two-step information-aware traffic reduction algorithm to address this problem. First, we provide a distributed and real-time algorithm for sensors to classify their measurements and report them selectively based on the importance of information. Then, we propose a bandwidth allocation algorithm to assign different forwarding probabilities to packets considering both the information quality and the network load. Our algorithm can be implemented and integrated easily with existing routing protocols to maximize the quality of information to the users, while reducing the amount of network traffic. We evaluate our approach based on the real sensing measurements to demonstrate the quality of information achieved. Simulations are also conducted to evaluate the performance of our proposed scheme in a larger network.