ATM networks have shown to reveal difficulties in simultaneously guaranleeing the requested Quality-of-Service while efficiently utilizing all the available bandwidth for economic reasons. The ABR (available bit rate) service class was designed to fill the gap of bandwidth with less delay hut loss sensitive traffic. Due to the bursty nature of higher-priority traffic and the resulting bandwidth fluctuations a flow control is required for ABR that avoids buffer overflow and cell loss. These demands can both only be supported by credit-based flow control. It offers immediate access to the full link bandwidth without a ramp-up time. However, though its behaviour is deterministic an analysis by queueing models is impossible. In this paper we treat stability, performance and dimensioning issues of a representative class of credit-based flow control protocols with a new stochastic Petri-Net model. We use its formalism to prove its stability and study the utilization performance in all operation regions with it. This implies the proposed dimensioning of buffer sizes and parameters, which is an important task during connection admission. IntroductionATM is the most promising technology for the next decades of networking. Its success is based on a fast packet-switched network technology supporting different service classes with specific Quality of Service (QoS) parameters and bandwidth demands that are guaranteed by the network. Constant and variable bit rate services support traffic of applications with precisely defined requirements for throughput and delay. The cell loss rate can however often he tolerated up to a certain amount. Thus statistical multiplexing [5] of hursty naffic is desired with a limited probability of cell loss due to buffer overflow. Nevertheless there always remains a considerable amount of spare bandwidth to ensue the low CLR and bounded delay [15]. The ABR service class has QoS requirements that are complementary to those of CBR and VBR There is no bandwidth and no delay guarantee, but cell loss is critical because of the kind of data that is transported by ABR. Its traffic is assumed to consist of data generated c.8. by FTP, HTTP, NFS or other computer communication protocols that are cmently based on TCP/IP in the Internet.If this traftic fills up the spare bandwidth in an uncontrolled way cell loss occurs at a switch when at the same time an output link is busy, the buffers are full and new cells arrive. Higher layer protocols react to cell loss with retransmissions leading to the congestion collapse phenomenon [l l] which must be avoided totally. This performance issue of cell loss for ABR is very important for customers, while link utilization is a major concern for providers.The goal of flow control on the ATM layer is to avoid buffer overflow by providing a closed loop feedback to sending nodes, instructing them to regulate their cell flow according to the semantics of the protocol. A flow control can only be effective and useful if each link is such a closed control loop, because its cont...

Corporate Alcatel-Lucent journal, Special Issue on Core and Wireless NetworksInternational audienceIn order to overcome current Internet limitations on overall network scalability and complexity, we introduce a new paradigm of semantic networking for the networks of the future, which brings together flow-based networking, traffic awareness, and self-management concepts to deliver plug-and-play networks. The natural traffic granularity is the flow between packet and circuit and between connection-less and connection-oriented modes. Using flow aggregation capabilities, we simplify traffic processing in the nodes through elastic fluid switching, and simplify traffic control through flow admission control, policing, and implicit quality of service (QoS) routing. By leveraging deep packet inspection and behavioral traffic analysis, network elements can autonomously and efficiently process the traffic flows they transport through real-time awareness gained via semantic analysis. The global consistency of node decisions within the whole network is ensured by self-management, applying the concepts of "knowledge plane" and "network mining.

No abstract

Abstract. With the introduction of new generation high speed routers, and with the help of "flow-aware" traffic management, it becomes possible to improve the Quality of Service for users as well as the network efficiency for ISPs. An example of the "flow-aware" traffic management is the Alcatel-Lucent "Semantic Networking" framework where short-lived flows are processed with high priority and long-lived flows are controlled on a per flow basis. In order to control efficiently the flows, it is useful to know an estimate of the Round Trip Time (RTT). In the present work, we provide an online RTT estimation algorithm which is passive and needs one-way traffic only. The one-way traffic requirement is essential for the application of the algorithm for "flow-aware" traffic management inside the network. To the best of our knowledge, there was no online one-way traffic RTT estimators. Tests on a controlled testbed and on the Internet demonstrate high accuracy of the proposed estimator.

No abstract

No abstract

Abstract. The phenomenal increase in network capacity to hundreds and thousands of Gbits/s in the core as well as Gbits/s at the access, is soon to witness stupendous amounts of packets that have to be processed and switched at amplifying line rates. Looking into the future, we address the need for the integration of packets of larger size, called XLFrames (XLFs), into the future Internet framework. This paper analyses the effects of introducing XLFs in a network that has both packets and XLFs. We evaluate the gains in terms of processing power and throughput. As we observe that XLFs have an impact on loss rate and fairness, we study how, with minimal efforts at routers while keeping the existing protocols (TCP/UDP, IP), XLFs may integrate in the current scenario.

Abstract-Looking into the future, this paper presents the effects of having packets of large sizes, called XLFrames (XLFs), in a network. The analysis is motivated by the fact that the Internet is soon to witness stupendous amounts of traffic that have to be processed and switched at amplifying line rates; and this brings forth multiple challenges in the form of energy efficiency, network performance and end-host performance. Increasing the size of packets in the Internet has far-reaching incentives that otherwise appear hard to achieve. We foresee an Internet that multiplexes both packets (sand) and XLFs (rocks). As a first step, we analyse the effects of introducing XLFs in a network, and find the following: (i) the amount of packet-header processing is greatly reduced, (ii) while the fair multiplexing of XLFs with standard packets can be achieved using a careful queue management in routers.