Hong-Yi Tzeng

Performance of TCP over UBR service in ATM networks with Per-VC early packet discard schemes

Both available bit rate (ABR) service and unspecified bit rate (UBR) service with early packet discard (EPD) schemes have been considered for supporting data applications in ATM networks. Since transmission control protocol (TCP) is perhaps the most widely used transport layer protocol in existing data networks, the performance of TCP over ATM using ABR service and UBR service with EPD schemes is of great interest to ATM equipment vendors and service providers. With regard to this interesting issue, we demonstrate in a companion paper [Li et al., 1996] some simulation results, which show that (1) with UBR, service and EPD schemes, TCP suffers significant performance degradation in terms of fairness and requires a relatively large switch buffer even with a small number of active virtual connections over a LAN configuration, and (2) for the same set of network configurations and with ABR service, TCP achieves good performance in terms of fairness and link utilization, and requires a relatively small switch buffer. In this paper, we extend our work in Li et al. by studying the performance of TCP over UBR service when EPD schemes are improved by using per-VC accounting and per-VC queueing techniques. In particular, our simulation results show that in a LAN environment, per-VC based EPD schemes can significantly improve the performance of TCP over UBR service in terms of fairness.

Congestion control for multicast service in ATM networks

In recent ATM Forum activities, considerable efforts have focused on the congestion control of point-to-point available-bit-rate (ABR) service. In this paper, we present a novel approach that extends the existing point-to-point (unicast) congestion control protocol to a point-to-multipoint (multicast) environment. In particular, we establish a unified framework to derive a multicast congestion control protocol for ABR service from a given rate-based unicast protocol. We first generalize a known necessary and sufficient condition on the max-min fairness of unicast rate allocation for multicast service. Then we show that the resulting multicast protocol derived using our framework preserves fairness characteristics of the underlying unicast protocol. The practical significance of our approach is illustrated by extending a standard congestion control mechanism for ABR service to a multicast environment. The performance of the resulting multicast protocol is examined using benchmark network configurations suggested by the traffic management subworking group at the ATM Forum, and simulation results are presented to substantiate our claims.

New dynamic SPT algorithm based on a ball-and-string model

A key functionality in todays widely used interior gateway routing protocols such as OSPF and IS-IS involves the computation of a shortest path tree (SPT). In many existing commercial routers, the computation of an SPT is done from scratch following changes in the link states of the network. As there may coexist multiple SPTs in a network with a set of given link states, such recomputation of an entire SPT not only is inefficient but also causes frequent unnecessary changes in the topology of an existing SPT and creates routing instability. This paper presents a new dynamic SPT algorithm that makes use of the structure of the previously computed SPT. This algorithm is derived by recasting the SPT problem into an optimization problem in a dual linear programming framework, which can also be interpreted using a ball-and-string model. In this model, the increase (or decrease) of an edge weight in the tree corresponds to the lengthening (or shortening) of a string. By stretching the strings until each node is attached to a tight string, the resulting topology of the model defines an (or multiple) SPT(s). By emulating the dynamics of the ball-and-string model, we can derive an efficient algorithm that propagates changes in distances to all affected nodes in a natural order and in a most economical way. Compared with existing results, our algorithm has the best-known performance in terms of computational complexity as well as minimum changes made to the topology of an SPT. Rigorous proofs for correctness of our algorithm and simulation results illustrating its complexity are also presented.

Message-optimal protocols for fault-tolerant broadcasts/multicasts in distributed systems with crash failures

An essential feature in any fault tolerant design of distributed systems is a mechanism by which a process can reliably broadcast information to other processes in the presence of failures. The paper studies the message complexity of fault tolerant broadcast protocols in weakly synchronous and totally asynchronous distributed systems with point to point communication links, where the system failures are caused by the processes but the communication links are completely reliable. We focus on the number of messages required of any fault tolerant protocol in failure free executions. Our motivation is that one should incur the cost of handling failures only when they actually occur. We present protocols that, in an n-process system subject to at most t crash failures where 1/spl les/t >

An efficient implementation of stack filters

Stack filters constitute a general class of nonlinear filters, including all order statistic filters such as median filters. Because of the stacking and threshold decomposition property, one can realize a stack filter as a sum of positive Boolean functions, which provides a simple VLSI implementation. The authors present a novel implementation of stack filters with complexity lower than previously developed architectures. In particular, the complexity of the implementation increases only linearly with window size.<<ETX>>

How effective are one-bit protocols?

Congestion control lies at the heart of the general problem of traffic management in asynchronous transfer mode (ATM) networks. The primary function of congestion control is to ensure good throughput/delay trade-off performance. This paper presents a fundamental study on the performance limits of one-bit protocols for congestion control-implemented by many ATM switch vendors because of its low hardware requirements-using a new complexity theoretic framework. We derive the first known tight bounds on the buffer size required for one-bit protocols that fully utilize link capacity and guarantee no cell loss under network congestion. In particular, we show that any such one-bit protocol will result in cN/sup 2/+O(dN), where N is the number of greedy sources in the network, d is the link delay, and the constant c is a parameter of the specific protocol.