The fundamental challenge in the traffic management of high-speed networks is to describe, analyse and control complex traffic flows in ways that can be applied in practice. Our approach to this problem is experimental in nature and starts with collection and analysis of high-resolution traffic traces from working high-speed networks. This paper analyzes cell-level ATM traffic and performance modeling issues of individual connections (i.e., VPI/VCIs) based on recent traffic measurements collected from a major North American public ATM network through a custom-built ATM traffic recording prototype device. In particular, we focus on the following issues related to modeling individual ATM connections at the cell level: the existence of ON-OFF behavior, sojourn time distributions of the ON and OFF periods, traffic arrival patterns within an ON period, temporal and spatial correlations, and performance impacts. Characteristics of individual connections can be highly dependent on the applications which generate the traffic and caution should be taken in interpreting these exact results for use in other ATM networks.

Recent traffic analyses from various packet networks have shown the existence of long-range dependence in bursty traffic. In evaluating its impact on queuing performance, earlier investigations have noted how the presence of long-range dependence, or a high value of the Hurst parameter H, is often associated with surprisingly large queue sizes. As a result, a common impression has been created of expecting queuing performance to be worse as H increases, but this impression can be misleading. In fact, there are examples in which larger values of H are associated with smaller queues. So the question is how can one tell whether queuing performance would improve or degrade as H rises? In this paper, we show that the relative queuing performance can be assessed by identifying a couple of time scales. First, in comparing a high-H process with a low-H process, there is a unique time scale tm at which the variances of the two processes match (assuming exact, second-order self similarity for both processes). Second, there are time scales tqi that are most relevant for queuing the arrivals of process i. If both of the queuing scales tqi exceed the variance-matching scale tm, then the high-H queue is worse; if the queuing scales are smaller, then the low-H queue is worse. However, no firm prediction can be made in the remaining case of tm falling between the two queuing scales. Numerical examples are given to demonstrate our results.

Packet traffic exhibits fractal or self-similar properties which are fundamentally different from features found in circuit switched voice traffic and captured by commonly used theoretical traffic models. These fractal properties are associated with the burstiness of packet traffic. In this paper we examine the impact of self-similarity on the link-layer error control mechanism, in particular, the most commonly used go-back-N automatic-repeat-request (ARQ) protocols. Our analysis indicates that in terms of the maximum throughput efficiency, the go-back-a protocol does achieve higher efficiency under the self-similar errors compared to the random error model. We also see that, assuming the underlying channel error process is self-similar, the Gilbert's 2-state error model does not predict the maximum throughput efficiency well; even with the parameter matching technique, we showed that the Gilbert's model only matches the maximum throughput efficiency well when the ratio of propagation delay to the frame transmission time is small. Self-similar traffic has higher delay compared to Poisson traffic for a given error model. For a given traffic stream, we see that the delay performance predicted by the conventional error models does not match the performance based on the self-similar error model. This suggests that conventional error models cannot capture the behavior of the go-back-N ARQ system performance. In general, with self-similar traffic and/or self-similar errors, network engineering needs to be more conservative than that based on conventional traffic and/or error models.

Motivated by field data which showed a large number of link changeovers and incidences of link oscillations between in-service and out-of-service states in common channel signalling (CCS) networks, a number of analyses of the link error monitoring procedures in the SS7 protocol were performed by the authors. This paper summarizes the results obtained thus far and include the following: (a) results of an exact analysis of the performance of the error monitoring procedures under both random and bursty errors; (b) a demonstration that there exists a range of error rates within which the error monitoring procedures of SS7 may induce frequent changeovers and changebacks; (c) an analysis of the performance of the SS7 level-2 transmission protocol to determine the tolerable error rates within which the delay requirements can be met; (d) a demonstration that the tolerable error rate depends strongly on various link and traffic characteristics, thereby implying that a single set of error monitor parameters will not work well in all situations; and (e) some recommendations on a customizable/adaptable scheme of error monitoring with a discussion on their implementability. These issues may be particularly relevant in the presence of anticipated increases in SS7 traffic due to widespread deployment of advanced intelligent network (AIN) and personal communications service (PCS) as well as for developing procedures for high-speed SS7 links currently under consideration by standards bodies.>

Link quality in the common channel signaling (CCS) network is monitored and reported by the Signaling System No.7 (SS7) protocol level 2 signaling link error monitoring procedures. The procedures include two algorithms: the signal unit error rate monitor (SUERM), which provides one of the criteria for taking the link out of service, and the alignment error rate monitor (AERM), which determines whether a link should be put into service. The authors, motivated by the excessive link status changes observed in some field operations of the CCS network provide a detailed analysis of the signaling link error monitoring algorithms and show that under current SS7 specifications of the error monitoring algorithms, the probability of a link oscillating in and out of service could be high for certain sustained error rates. While the risk to the existing network may not be high due to the light loads carried at present, there is still a need to study the error monitoring issues thoroughly.>

The load splitting link selection algorithm used in the Common Channel Signaling (CCS) network and proposed for use in the interswitching system interface in the network supporting the Switched Multimegabit Data Service (SMDS) is considered. The algorithm uses a randomly generated bit pattern to select a link for transmission. The author discusses how the traffic loads are distributed among the links in a link set based on the load splitting link selection algorithm and shows that, in some cases, the overall traffic capacity of the link set does not increase even when the number of installed links increases. A theorem that shows how to construct the set of links to increase the line set capacity is demonstrated. The theorem is used as the basis for an efficient and flexible procedure for implementing the link dimensioning algorithm in software, to support the mechanization of the CCS and SMDS network planning and traffic engineering functions.>

Methods for determining the network reliability of a multihop packet radio network in the presence of hostile jammers are discussed. A new connectivity parameter called radio connectivity is defined as the maximum number of disjoint communication paths that are still usable between given nodes s and d after the jammer is on or, more generally, the minimum number of jammers needed to disconnect s and d. A lower bound on the radio connectivity is computed by studying the number of jamming independent paths. The time complexity of obtaining the radio connectivity is analyzed and shown to be NP-hard except for some special cases. Greedy heuristics for developing approximate answers for general networks are described. Euclidean networks, in which the nodes and links correspond to points and line segments in the Euclidean geometry and satisfy Euclid's four fundamental axioms, are also discussed. It is found that the maximum number of independent paths between a pair of source and destination nodes that can possibly exist is five. An extension in which there is a protected zone of known size around the sender and receiver is studied.>

The authors study data link error control protocols suitable for point-to-multipoint communication where data are delivered to the destinations in the order they are sent. They study a series of adaptive protocols differing in the way that the sender uses the outcomes of the previous transmissions. The protocols are based on the go-back-N schemes discussed by I.S. Gopal and J.M. Jaffe (1984), whose protocols are generalized to the case where multiple copies of a message are sent (instead of just a single copy). The optimum number of copies is determined and is found to depend on how many receivers have not yet received the message. These adaptive protocols can be applied to all three basic ARQ (automatic repeat request) schemes, namely, stop-and-wait, go-back-N, and selective-repeat. A dynamic programming technique is used to solve this optimization problem. A throughput comparison shows that, by sending the optimum number of copies of a data frame instead of just a single copy, the performance will be significantly improved.>

Motivated by the excessive link status changes observed in some field operations of the common channel signaling (CCS) network, the authors provide a detailed analysis of the signaling link error monitoring algorithms in the Signaling System No. 7 (SS7) protocol. These algorithms determine when to fail a link due to excessive error rates and when to put a failed link back into service. The analysis shows that, under current SS7 specifications of the error monitoring algorithms, the probability of a signaling link oscillating in and out of service could be high, depending on the traffic load, signal unit size, and the statistical nature of errors (bursty or random). The link oscillation phenomenon could become worse as longer Transaction Capability Application Part (TCAP) messages for transaction-based services (e.g., 800 Service) are carried in the CCS networks. While the risk to the existing network may not be high due to the light loads carried at present, there is still a need to study the error monitoring issues thoroughly.>

The integrated network planning system (INPLANS) is introduced. It was developed to provide a single environment to support regional network planning and traffic engineering across different networking technologies. There are four main capabilities in INPLANS: integrated network monitoring (INM), which supports planning and engineering monitoring studies; integrated network servicing (INS), which provides an integrated and interactive environment for the network servicing process; integrated capacity expansion (ICE), which provides a 1-5 year view of network demands and requirements; and integrated technology planning (ITP), which provides cost effective backbone and network access design plans. The overall network planning and traffic engineering process and how it is supported and implemented in INPLANS are described. The switched multimegabit data service (SMDS) is considered, and the impact of SMDS on the planning and engineering process is described.>

Some data-link layer error control go-back-N ARQ (automatic repeat-request) protocols are studied that are suitable for point-to-multipoint communication over broadcast channels where data are delivered to the destinations in the order they are sent. A series of protocols differing in the way that the sender uses the outcomes of the previous transmissions are studied. The system delay, rather than the throughput, is the optimization measure. The optimal number of copies that the sender should transmit to minimize the time between when the sender first transmits a data frame and when the data frame is accepted by all the receivers is determined. The results show that sending the optimum number of copies of a data frame instead of just a single copy significantly improved the delay performance.>

A study is made of some data-link-layer error-control protocols suitable for point-to-multipoint communication over broadcast channels where data are delivered to the destinations in the order they are sent. The authors study a series of protocols differing in the way that the sender uses the outcomes of the previous transmissions. The protocols are based on the go-back-N schemes discussed by I.S. Gopal and J.M. Jaffe (1984). The authors generalize their proposed protocols to the case where multiple copies of a message are sent (instead of just a single copy). The optimum number of copies is determined, which depends not only on the round-trip propagation delay of the channel and the error probability, but also on the number of receivers that have not yet received the message. A dynamic programming technique is used to solve this optimization problem. The throughput comparison shows that by sending the optimum number of copies of a data frame instead of just a single copy, the performance will be significantly improved.>