Levent Gun

Bandwidth management and congestion control framework of the broadband network architecture

Abstract In this paper we describe an integrated set of procedures for bandwidth management and congestion control in IBMs broadband network architecture for high speed networks. These controls operate at two different time scales, at the connection level and at the packet or cell level. Major connection level controls are path selection, admission control and bandwidth allocation. These controls are applied at the connection setup time and are based on the connection characterization and the network state at that time. Packet level controls are access control, traffic monitoring and intermediate node buffer management. Because of their smaller time scale, they do not rely on any feedback information from the network. The main focus of this paoer is to reconcile the different time scales of these controls, and define how they interplay to ensure proper network operation. The ideas presented in this paper are being implemented and deployed in several test-beds and field trials for experimental validation.

Parallel queues with resequencing

This paper considers a system where Poisson arrivals are allocated to K parallel single server queues by a Bernoulli process. Jobs are required to leave the system in their order of arrival. Therefore, after its sojourn time T in a queue a job also experiences a resequencing delay R, so that the time in system for a job is S = T + R. The distribution functions and the first moments of T, R, and S are first obtained by sample path arguments. The sojourn time T is shown to be convex in the load allocation vector in a strong stochastic sense defined in [21]. It is also shown that, in a homogeneous system, equal load allocation minimizes both the random variable T (in the usual stochastic order) and the system time S (in the increasing convex order)

NBBS traffic management overview

In this paper, we describe an integrated set of procedures used for bandwidth management and congestion control in high-speed packet-switched networks such as asynchronous transfer mode (ATM), which are part of IBMs Networking BroadBand Services (NBBS) architecture. These controls are designed to support a wide variety of services with different characteristics in the network and operate at different time scales: connection-level controls such as path selection, admission control and bandwidth allocation, and packet-level controls that discriminate between packets from different connections to support multiple levels of service guarantees. Connection-level controls are applied at connection setup time and are based on the connection characterization and the network state at that time. They perform efficient allocation of resources to ensure performance guarantees for connections while achieving high utilization of network resources. Various packet-level controls developed include access or rate control and intermediate node buffer management and scheduling. For connections that do not require explicit service guarantees, NBBS offers an available bit rate service. This service mostly relies on packet-level control in the form of an end-to-end rate-based flow control algorithm that regulates the flow of traffic into the network. This paper, in addition to providing an overview of the different mechanisms used for traffic management in NBBS, highlights how they interact to ensure efficient network operation.

Effectiveness of dynamic bandwidth management mechanisms in ATM networks

The leaky bucket, a credit management system that allows cells to enter the network only if sufficient credit is available, is considered. Motivation for an in-depth study of dynamic control algorithms, to avoid congestion in high-speed networks under time varying traffic statistics, is provided. The need for real-time traffic measurements and dynamic control actions for long-lived connections is demonstrated. In particular, it is shown through simulations that implementing access control algorithms based on the bandwidth allocation procedures only at connection setup is not sufficient for congestion-free operation of the network. It is further shown that if the bandwidth allocation and the leaky bucket parameters are not dynamically adjusted, the performances of the connections depend on their initial parameters and can be very undesirable as the traffic parameters change over time. Moreover, if corrective actions are not taken, congestion may build up in the network even in the presence of leaky buckets. The simple dynamic approach proposed limits the congestion periods in the network to the time scales of the connection-level controls. Therefore, the probability of many connections sending excess traffic into the network at the same time is greatly reduced, alleviating the congestion problem.<<ETX>>

NBBS path selection framework

This paper describes the path selection function in Networking BroadBand Services (NBBS), which is IBMs architecture for high-speed, multimedia networks. The distinguishing feature of a multimedia network is its ability to integrate different applications with different traffic characteristics and service requirements in the network, such as voice, video, and data. In order to meet their service requirements, it is necessary for the network to provide unique quality-of-service (QOS) guarantees to each application. QOS guarantees, specified as multiple end-to-end performance objectives, translate into path and link constraints in the shortest path routing problem. For a general cost function, shortest path routing subject to path constraints is known to be a nonpolynomial- (NP-) complete problem. The NBBS path selection algorithm, a heuristic solution based on the Bellman-Ford algorithm, has a polynomial order of complexity. The algorithm finds a minimum hop path satisfying an end-to-end delay (or delay variation) constraint, that in most cases also optimizes a load balancing function. To reduce the number of path constraints, other QOS requirements such as packet loss ratio are implemented as a link constraint. The notion of primary and secondary links is used to minimize the long-term overall call blocking probability by dynamically limiting the hop count of a given path. The path selection algorithm developed for point-to-point connections is described first, followed by its extension to the case of point-to-multipoint connections.

Bandwidth allocation and access control in high-speed networks

The problem of bandwidth allocation and access regulation arises in the congestion control of Broadband ISDN networks. This paper assumes that a single user, described by an on-off fluid model, is connected to the network via a leaky bucket access control mechanism. The bandwidth allocated to this user and the leaky bucket parameters are to be selected so as to guarantee a negotiated level of delay probability at the access point and packet loss probability in the network which is modelled as an output buffer. The design problem is to minimize the allocated bandwidth subject to service guarantees and stability conditions for the input and output buffers. We provide a desirable feasible solution to the design problem. The paper studies the effect of non-conforming users on the network performance using the leaky bucket access control corresponding to this feasible solution. We provide expressions that quantify the impact of the leaky bucket parameters in access regulation and the worst-case queueing behavior at the output buffer. Finally, we discuss the extension of this methodology to the multiple leaky buckets case.

Effective bandwidth vector for two-priority ATM traffic

Traffic consisting of two classes (0 and 1) arrives at a buffer in a bursty fashion. The type-0 traffic is always admitted into the buffer (of size B/sub 0/) as long as there is space for it. Type-1 traffic is admitted to the buffer only if the buffer content does not exceed a threshold B/sub 1spl les/B/sub 0/. The authors develop an analytical model to determine the bandwidth allocation in order to ensure that the packet loss probability of type-i traffic does not exceed /spl epsisub i/, where 0</spl epsisub 0spl lesspl epsisub 1/<1 are given numbers. These results extend the notion of effective bandwidth of a single class traffic. They also studied the variation of this effective bandwidth with the buffer threshold and with the amount of high priority traffic. This study provides valuable insights into the buffer dimensioning problem as well as the amount of bandwidth that can be saved by using priority discard mechanisms.<<ETX>>

An approximation method for capturing complex traffic behavior in high speed networks

This paper provides a simple technique to approximately map complex user traffic behavior to a simple traffic model which can be fully characterized by three parameters. This approximate characterization relies on simple measurements taken at the access point to the network and enables various bandwidth management procedures being implemented for high speed networks (see Cidon et al., 1991; Gu6rin and Grin, 1992; Giin and Gufrin, 1993) to be applicable for very general classes of user traffic. Through numerical examples, we show that this simple approach provides sufficiently accurate estimates of the potential impact of the true process on the network. Applications to characterizing multiplexed processes inside the network are also discussed.Abstract This paper provides a simple technique to approximately map complex user traffic behavior to a simple traffic model which can be fully characterized by three parameters. This approximate characterization relies on simple measurements taken at the access point to the network and enables various bandwidth management procedures being implemented for high speed networks (see Cidon et al., 1991; Guerin and Gun, 1992; Gun and Guerin, 1993) to be applicable for very general classes of user traffic. Through numerical examples, we show that this simple approach provides sufficiently accurate estimates of the potential impact of the true process on the network. Applications to characterizing multiplexed processes inside the network are also discussed.

An overview of bandwidth management procedures in high-speed networks

Packet-switched networks such as ATM [1] and plaNET [2] will form the basis for multimedia high speed networks. In these networks traffic streams with widely varying characteristics such as voice, data and video will be statistically multiplexed and share common switching and transmission resources. Managing the available bandwidth to avoid congestion and provide guaranteed levels of Grades of Service (GOS) for connections with potentially dramatic differences in their statistical behavior pose new challenges very different from the ones present in traditional packet- or circuit-switched networks.