Yung-Terng Wang

Analysis of cyclic service systems with limited service: bounds and approximations

Abstract In this paper we analyze cyclic service queueing systems in which the server serves at most a prespecified number of jobs when it visits a queue. The system parameters are not required to be symmetric. Both exhaustive-limited and gated-limited disciplines are studied, which include well-known disciplines such as exhaustive, gated and serve-at-most-one as special cases. We derive tight bounds for the conservation laws in such systems. These bounds reduce to exact equalities for several known special cases. Then, simple approximations are proposed for the mean waiting times of jobs at each queue in such systems. Extensive numerical examples indicate that the accuracy of the approximations is well within 20% of the simulated results provided they are not extremely asymmetric and the switchover times are not large relative to the service times.

First-order rate-based flow control with dynamic queue threshold for high-speed wide-area ATM networks

In this paper we present a new rate-based flow control scheme for ATM ABR services and analyze its performance. The proposed algorithm, which we refer to as first-order rate-based flow control (FRFC) is the most simple form of queue-length-based flow control. The asymptotic stability, the steady-state throughput, queue length and fairness, and the transient behavior are analyzed for the case of multiple connections with diverse round-trip delays. We also consider a novel approach to dynamically adjust a queue threshold in the FRFC according to the changes in the available bandwidth, and the arrival and departure of connections. Simulations show that the simple FRFC with dynamic queue threshold (DQT) effectively maintains high throughput, small loss and a desired fairness in these dynamic environments and is a promising solution for ABR flow control in ATM networks.

Designing stable ABR flow control with rate feedback and open-loop control: first-order control case

In this paper we present a control-theoretic approach to design stable rate-based flow control for ATM ABR services. The flow control algorithm that we consider has the most simple form among all the queue-length-based flow control algorithms, and is referred to as first-order rate-based flow control (FRFC) since the corresponding closed loop can be modeled as a first-order retarded differential equation. We analyze the equilibrium and the asymptotic stability of the closed loop for the case of multiple connections with diverse round-trip delays. We also characterize the asymptotic decay rate at which the stable closed loop tends to the equilibrium. The decay rate is shown to be a concave function of control gain with its maximum being the inverse of round-trip delay. We also consider an open loop control in which the queue control threshold is dynamically adjusted according to the changes in the available bandwidth and the number of connections. This open loop control is shown to be necessary and effective to prevent the closed loop from converging to an undesirable equilibrium point.

VoIP network architectures and QoS strategy

Voice over IP (VoIP) has received much attention in recent years with the promise of lower costs, as well as new revenue-generating services. Cost and services advantages of carrying voice over IP compared to over the current circuit network are made possible by a common high-capacity packet infrastructure for voice, data, and multimedia services. An important requirement of such a packet infrastructure is the ability to provide public-switched telephone network (PSTN) grade quality without excessive over-provisioning. In this paper, we describe an approach to offer AbsoluteQoS™ to voice and other demanding applications over a general-purpose packet network. AbsoluteQoS is defined as the ability to provide an engineered bound on call-blocking and quantitative QoS guarantee that calls-in-progress will receive. The proposed strategy is based on key innovations in architectures and protocols, as well as business models of PSTN and packet networks.

Integrated Network Design Tools (INDT): a suite of network design tools for current and next generation networking technologies

Integrated Network Design Tools (INDT) is a suite of tools being developed at Bell Labs. This suite of tools is built on a flexible software architecture based on C++ and can provide optimal network design for a mix of private line, switched voice and switched multimedia services using a mix of circuit and ATM switches and cross connects and facilities with different bandwidth granularity. INDT also has the ability to design, in a standalone mode or in conjunction with logical designs, the core transport network using mesh and/or ring topologies and using PDH or SONET/SDH products. The transport network may be designed over existing fiber routes and capacities or the tool may be used to help decide the fiber layout. We focus on a new capability within INDT, viz., SONET ring design. We present a high level description of the various submodules in the ring network design and provide an overview of the main factors that influenced our algorithm development. We also present sample studies of our approach to intra-ring load balancing as well as our economic quantification of penalties due to dual ring interconnect (DRI) and other features of interest.

An ABR rate-based congestion control algorithm for ATM switches with per-VC queueing

The support of available bit rate (ABR) service in ATM networks has been extensively studied in literature under the assumption that all ABR traffic is queued in the switches in a common first-in-first-out (FIFO) queue. In this context, which models what is typically done in current generation switches, the central issue has been the design of ABR schemes for the switch behavior that are able to allocate an accurate fair share of available bandwidth to each ABR virtual connection (VC). The new generation of switches that will soon be deployed, however, provides per-VC queueing and scheduling to multiplex traffic from different VCs. In this paper, we show how the ABR scheme used in the switches can take advantage of the new queueing and scheduling capabilities. Since the per-VC scheduler itself is able to provide flow isolation and fair service to the VCs, the ABR scheme is not required to accurately compute the fair share, but can focus primarily on queue control. Thus, a simple ABR scheme that only uses an approximation of the fair share is sufficient to achieve excellent performance. Following these principles, we present a new algorithm, called the enhanced dynamic max rate control algorithm (EDMRCA), which is the extension to per-VC queueing of DMRCA, a simple queue-length-based algorithm developed for common-FIFO queueing. EDMRCA does not suffer the problems of its common-FIFO counterpart in allocating the fair share when ABR traffic interferes with highly-bursty traffic and in other demanding traffic scenarios, and inherits the robustness and excellent congestion control properties of DMRCA. EDMRCA achieves fairness at least comparable with schemes using accurate computation of the fair share, offers superior queue control, and requires only a fraction of the complexity.

Architecture and performance of a multi-Tbps protocol independent switching fabric

We describe a scalable multi-Tbps protocol independent switching fabric architecture. This architecture is cost effective for building switches supporting wirespeed switching up to 1024 ports with capacity ranging from 10s of Gbps to 10s of Tbps and QoS traffic classes. The scalability has been achieved via self-routing of cells with the routing queue number scheme coupled with an efficient backpressure and flow control. We also showed that arbitrarily high utilization of the crossbars can be achieved using a simple output buffering and re-sequencing scheme.

A conservation law based approximate analysis for a class of simultaneous resource possession problems

Abstract In this paper we analyze a single server queueing model in which there are two types of jobs, one of which must wait in an external queue until a token is available, and only then may join the service queue. The interarrival times and service requirements for both types of jobs are assumed to be independent and exponentially distributed. We derive the stability condition for such a model where the service queue discipline is either FCFS (First-Come-First-Serve) or PS (Processor-Sharing). We then propose analytic approximations for the mean waiting times for both types of jobs, relying heavily on the M/G/1 conservation law. Numerical results show that our approximations are very accurate (within a few percent of the simulated results) even when the system is heavily loaded. The approximations are also shown to be asymptotically exact as the number of tokens N → ∞.

A unified ABR flow control approach for multiple-stage ingress/egress-queueing ATM switches

Most of the existing ATM available bit rate (ABR) flow control algorithms are designed based on the assumption of a simple output-buffered switch architecture. Under such an assumption, congestion can only occur at the output ports of the switch. Moreover, multiple output ports of an ATM switch can be modeled as independent queues whose congestions are independent of each other. Previously, however, research/commercial ATM switches have been evolving towards a multiple-stage architecture which contains both input and output queueing to improve capacity scalability. With the new architectures, congestion can develop at different locations within a switch. More importantly, the onset of these congestions may dependent on each other. It therefore becomes necessary for any ABR flow control algorithm to handle multiple dependent bottlenecks under such architecture. In this paper, we describe a unified ABR flow control strategy for the new generation ATM switches. Our design is geared towards the multi-stage, input/output-queueing architecture. Our strategy can be used to adapt any existing output-buffering focused, queue-length based ABR algorithm for the new architecture. As a concrete example, we discuss the adaptation of the DMRCA ABR flow control algorithm for a multiple-stage input/output queueing switch. The result is a utilization-based adaptive dual DMRCA algorithm which achieves (1) low cell-loss, (2) high switch utilization and (3) per-VC fairness in bandwidth allocation for VCs across multiple ingress buffers when traffic patterns permit. We also present results from simulation studies.

Flow control in a high-speed bus-based ATM switching hub

Studies flow control in a high-speed bus-based ATM switching hub for premises switching. The switching fabric is a dual-bus with slots for diverse port cards that interface to the external world. There can be a significant discrepancy in the switching fabric speed and the port card speed. This can result in buffer overflows at the receiving port buffer and consequently high losses in the switching fabric. Adequate flow control mechanisms are necessary. This paper first examines two different flow control strategies. Then we consider a third hybrid strategy which combines the strengths of the first two strategies. In the first flow control scheme, physical flow control, all streams destined to a receiver with buffer problems such as high occupancy and loss are shut down till the congestion is cleared. This scheme has the advantage that loss is severely limited but has the disadvantage that high-rate streams arriving at this buffer can in some circumstances starve lower-rate streams. In a second strategy, logical flow control, streams are implicitly selected based on their rates and shut down. This scheme has the advantage that the higher-rate streams will eventually be shut down more often and hence cannot overwhelm the lower-rate streams. However, loss is not easily controlled in this scheme. Finally, in a hybrid strategy, we combine physical and logical flow control with logical control activated first and physical control activated later when the logical control is unable to limit the buffer occupancy and loss. This hybrid strategy is the recommended choice for flow control in the setting of interest.