TCP Fluid Modeling with a Variable Capacity Bottleneck Link (original) (raw)

A fluid-based model of time-limited TCP flows

Computer Networks, 2004

Design methodologies for TCP/IP based networks is one of the most challenging topic in research on telecommunications networks. The main difficulty consists in modeling congestion control mechanisms because it involves feedback from the network. In this perspective, in this paper we develop an accurate analytical framework for networked TCP applications supporting both Slow Start and Congestion Avoidance algorithms. To this end we enhance the TCP fluid model introduced in the previous literature considering also TCP sources aimed at transmitting a prefixed quantity of data such as file transfer or network browsing. The proposed model addresses a network of routers supporting any Active Queue Management (AQM) techniques, provided that the equations describing the AQM rules implemented in the routers are introduced in the fluid framework. The proposed framework allows designers to study not only the steady-state behavior of the network, but also the transient behavior when a set of TCP sources start to transmit or finish transmitting. Moreover the model results provide the average values of the queue length in each router even when some of them are not bottleneck ones.

TCp throughput and timeout - steady state and time-varying dynamics

IEEE Global Telecommunications Conference, 2004. GLOBECOM '04.

While many models of the TCP's dynamics have been developed, few focus on the effects of timeout and high loss probability. Active queue management (AQM) is an important application of these dynamic models. However, recent work has shown that AQM provides little performance benefit over drop-tail queueing for HTTP traffic, except possibly at high utilizations. It is at these utilizations that the dynamic models of TCP are the least accurate. This paper presents a dynamic model of TCP that accurately models timeout. This model is also applicable to the static case. This paper also presents a model of the variance and the distribution of the congestion window. It is shown that, while the dynamics of the mean value of the congestion window are rather tame, the dynamics of timeout has large oscillations that take several seconds to decay. These oscillations cause the average bit-rate to also wildly oscillate. Finally, this paper includes results from several million simulations providing a detailed view of the dynamics of timeout.

Pitfalls in the fluid modeling of RTT variations in window-based congestion control

2005

Deterministic delay differential equation models, where the packet traffic is modeled as a fluid, are widely used to study congestion control algorithms in the Internet. In this paper, we point out some pitfalls in such fluid modeling of window flow control algorithms. Specifically, we argue that the modeling assumptions used to capture the variability in the RTT (due to queue length fluctuations) may play a critical role in our ability to design stable algorithms. We study two scenarios to illustrate the dramatic impact of RTT modeling. We first consider TCP-Reno with RED, and show that assuming that the RTT is a constant (when it is actually time-varying) leads to conservative parameter choices, i.e., the system continues to be stable even with variable RTT. On the other hand, for the recently proposed Stabilized Vegas, we show the following result: while the network can be stabilized under the constant RTT assumption, there is no choice of parameters that would stabilize the system when the RTT variations are taken into account! Interestingly, such problems do not arise if the congestion-control mechanisms at the end-users are rate-based.

A model for TCP congestion control capturing the correlations in times between the congestion events

2006 2nd Conference on Next Generation Internet Design and Engineering, 2006. NGI '06., 2006

We consider a simplified model for the rate control of TCP sources. In particular, we assume idealized negative feedbacks upon reaching a certain total sending rate, i.e., at the moment when the total sending rate attains a given capacity limit c one of the TCP sources is given a negative feedback and the source reduces its sending rate in a multiplicative manner. Thus, the model takes into account the interactions between different flows appropriately at the microscopic level instead of assuming independence. For this model we are able to derive steady state equations and solve them. Furthermore, we are able to compute several important performance measures such as the mean and the variance of the total sending rate. Moreover, we are able to characterize the packet loss process at the bottleneck link and, in particular, the correlations therein.

Improved Fluid-Model of TCP/AQM Network for Congestion Control

The Open Automation and Control Systems Journal, 2009

In this paper, recently proposed active queue management (AQM) algorithms for supporting end-to-end transmission control protocol (TCP) congestion control is revisited. We focus recently developed theoretic results on design and analysis for the AQM based TCP congestion control dynamics. In this context, the existing fluid model of the TCP/AQM network is discussed. Moreover, an improved fluid model is addressed, taking time delay in inner feedback loop into account, which is neglected in the modeling process of fluid model. The stabilization of the fluid model is investigated, which has shown that the stabilizing region of PID congestion controller for the conventional fluid model moves 1 k and 1 T k along the p k and d k axes respectively compared with that for the improved fluid model, ie. the actual stability region. The improved fluid model has a great potential in analyzing and designing various network congestion control algorithms.

Modeling the interaction between TCP and rate adaptation at links

Proceedings of the 2013 25th International Teletraffic Congress (ITC), 2013

In this paper, we model and investigate the interaction between the TCP protocol and rate adaptation at intermediate routers. Rate adaptation aims at saving energy by controlling the offered capacity of links and adapting it to the amount of traffic. However, when TCP is used at the transport layer, the control loop of rate adaptation and one of the TCP congestion control mechanism might interact and disturb each other, compromising throughput and Quality of Service (QoS). Our investigation is lead through mathematical modeling consisting in depicting the behavior of TCP and of rate adaption through a set of Delay Differential Equations (DDEs). The model is validated against simulation results and it is shown to be accurate. The results of the sensitivity analysis of the system performance to control parameters show that rate adaptation can be effective but a careful parameter setting is needed to avoid undesired disruptive interaction among controllers at different levels, that impair QoS.

Observing the effect of TCP congestion control on network traffic

In this paper we quantitatively observe the relationship between packet loss rate and degree of traffic burstiness using ns-2 network simulations. We examine the traffic behavior of a single TCP connection in a lossy environment. The loss agent in the simulation uniformly drops incoming packets with a given probability. By changing the packet loss rate, we can vary the congestion level of network and influence the behavior of TCP's congestion control mechanism. We measure the variation of traffic burstiness as the packet loss rate is changed. We also observe different traffic behavior with respect to the types of traffic sources. It is observed that for exponential traffic source, the degree of traffic burstiness is influenced by the packet loss rate over some range. However, the degree of traffic burstiness for Pareto traffic source is almost the same regardless of the packet loss rate. While our simulation is similar to previous works [1][2], the observation is interesting since the relationship between packet loss rate and degree of traffic burstiness is measured quantitatively.

Analysis of Some Deterministic Models for TCP Congestion Window Control

2001

This paper discusses some results on the Transport Control Protocol (TCP) congestion window control, based on experiments using Network Simulators, and also on Steady-State Deterministic Models. Despite the numerous observations that the traffic control exhibits periodic and even chaotic behaviour for different service rates and for different buffer sizes, there still is a lack of analytical models describing its dynamical behaviour. To palliate this problem, an iterative model, analogous to the one developed in the frame of the max-plus algebra has been selected and presented. It is shown that this model fits the dynamics of the TCP congestion window control the case of additive increase, multiplicative decrease (AIMD) mechanism, but also in the case of AIMD(a,b) mechanism. This model could be further adapted to optimise the congestion

TCP modeling in the presence of nonlinear window growth

2001

Abstract: We develop a model for TCP that accounts for both sublinearity and limitation of window increase. Sublinear window growth is observed when the round-trip time of the connection increases with the window size. The limitation is due to the window advertised by the receiver. First, we derive the required conditions for the stability of the model. Then, we write the Kolmogorov equation under Markovian assumptions. The model is solved 1analytically for some particular cases. A good match between the throughput predicted ...

Beyond fluid models: Modelling TCP mice in IP networks under non-stationary random traffic

Computer Networks, 2007

Fluid models of IP networks are based on a set of ordinary differential equations, that provide an abstract deterministic description of the average network dynamics. When IP networks operate close to saturation, fluid models were proved to provide reliable performance estimates. Instead, when the network load is well below saturation, standard fluid models lead to wrong performance predictions, since all buffers are forecasted to be always empty, so that the packet discard probability is predicted to be zero. These incorrect predictions are due to the fact that fluid models, being deterministic in nature, do not account for the random traffic variations that may induce temporary congestion of some network elements. In this paper we discuss three different approaches to describe random traffic variations in fluid models, considering randomness at both the flow and packet levels. With these approaches, fluid models allow reliable results to be obtained also in the case of IP networks that operate well below their saturation load. Numerical results are presented to prove the accuracy and the versatility of the proposed approaches, considering both stationary and non-stationary traffic regimes.