Victor Firoiu

Modeling TCP Reno performance: a simple model and its empirical validation

The steady-state performance of a bulk transfer TCP flow (i.e., a flow with a large amount of data to send, such as FTP transfers) may be characterized by the send rate, which is the amount of data sent by the sender in unit time. In this paper we develop a simple analytic characterization of the steady-state send rate as a function of loss rate and round trip time (RTT) for a bulk transfer TCP flow. Unlike the models of Lakshman and Madhow (see IEE/ACM Trans. Networking, vol.5, p.336-50, 1997), Mahdavi and Floyd (1997), Mathis, Semke, Mahdavi and Ott (see Comput. Commun. Rev., vol.27, no.3, 1997) and by by Ott et al., our model captures not only the behavior of the fast retransmit mechanism but also the effect of the time-out mechanism. Our measurements suggest that this latter behavior is important from a modeling perspective, as almost all of our TCP traces contained more time-out events than fast retransmit events. Our measurements demonstrate that our model is able to more accurately predict TCP send rate and is accurate over a wider range of loss rates. We also present a simple extension of our model to compute the throughput of a bulk transfer TCP flow, which is defined as the amount of data received by the receiver in unit time.

A study of active queue management for congestion control

In this work, we investigate mechanisms for Internet congestion control in general, and random early detection (RED) in particular. We first study the current proposals for RED implementation and identify several structural problems such as producing large traffic oscillations and introducing unnecessary overhead in the fast path forwarding. We model RED as a feedback control system and discover fundamental laws governing the traffic dynamics in TCP/IP networks. Based on this understanding, we derive a set of recommendations for the architecture and implementation of congestion control modules in routers, such as RED.

On achievable service differentiation with token bucket marking for TCP

The Differentiated services (diffserv) architecture has been proposed as a scalable solution for providing service differentiation among flows without any per-flow buffer management inside the core of the network. It has been advocated that it is feasible to provide service differentiation among a set of flows by choosing an appropriate “marking profile” for each flow. In this paper, we examine (i) whether it is possible to provide service differentiation among a set of TCP flows by choosing appropriate marking profiles for each flow, (ii) under what circumstances, the marking profiles are able to influence the service that a TCP flow receives, and, (iii) how to choose a correct profile to achieve a given service level. We derive a simple, and yet accurate, analytical model for determining the achieved rate of a TCP flow when edge-routers use “token bucket” packet marking and core-routers use active queue management for preferential packet dropping. From our study, we observe three important results: (i) the achieved rate is not proportional to the assured rate, (ii) it is not always possible to achieve the assured rate and, (iii) there exist ranges of values of the achieved rate for which token bucket parameters have no influence. We find that it is not easy to regulate the service level achieved by a TCP flow by solely setting the profile parameters. In addition, we derive conditions that determine when the bucket size influences the achieved rate, and rates that can be achieved and those that cannot. Our study provides insight for choosing appropriate token bucket parameters for the achievable rates.

Performance evaluation of ATM shortcut connections in overlaid IP/ATM networks

In this paper we present methods to evaluate the benefit of using direct ATM connections (shortcuts) between IP nodes in IP over ATM networks, and we identify the combinations of IP and ATM network topologies where ATM shortcut benefits are likely to be high. We model an IP/ATM network with and without ATM shortcuts as two loss networks. We propose a metric for network performance comparison, the network load ratio, that gives the ratio of the number of flows accepted by two networks at the same network blocking probability. We derive an estimator of this metric, the asymptotic load ratio, that has low computational complexity. This estimator forms the basis of a methodology for network performance comparison. We use this method in simulation experiments using random networks. These experiments indicate that in many cases the utilization of an IP/ATM network increases proportionally to the decrease in the average path length when ATM shortcuts are used. We have also found that there is almost no correlation between the increase in network utilization (when using ATM shortcuts) and the IP to ATM node ratio.

Feedback output queuing: a novel architecture for efficient switching systems

We consider the problem of providing QoS guarantees in a high-speed packet switch. As basic requirements, the switch should be scalable to high speeds per port, a large number of ports and a large number of traffic flows with independent QoS guarantees. Existing scalable solutions are based on virtual output queueing, which is computationally complex when required to provide QoS guarantees for a large number of flows. We present a novel architecture for packet switching that provides support for such QoS guarantees. A cost effective fabric with small external speedup is combined with a feedback mechanism that enables the fabric to be virtually lossless, thus avoiding packet drops indiscriminate of QoS flows. Through analysis and simulation, we show that this architecture provides accurate QoS support, has low computational complexity and is scalable to very high port speeds.

Traffic models and admission control for variable bit rate continuous media transmission with deterministic service

Abstract In this paper we address the problem of call admission based on general descriptions of stored continuous media (CM) flows. We show that such an approach to call admission can result in a very high resource utilization that can be significantly higher than that produced when using time-invariant flow descriptions. We present the admissibility conditions for flows, where packets are scheduled according to the earliest-deadline-first (EDF) scheduling policy. We further present an algorithm for testing for admissibility of a new flow whose computational complexity is linear in the number of flows and in the linear segments used for flow description. Last, we present an algorithm for producing a parsimonious flow description which produces most of the benefit of the best general description while requiring a small number of linear segments. Simulation shows that this improves the network resource utilization as much as 200–250% over the best possible utilization achieved using any time-invariant workload function. A significant improvement results even when the workload is smoothed.