Mario Barbera

Active Window Management: An Efficient Gateway Mechanism for TCP Traffic Control

Many techniques have been proposed in the last few years to address performance degradations in end-to-end congestion control. Although these techniques require parameter tuning to operate in different congestion scenarios, they miss the challenging target of both minimizing network delay and keeping goodput close to the network capacity. In this paper we propose a new mechanism, called Active Window Management (AWM), which addresses these targets by stabilizing the queue length in the network gateways. AWM acts on the Advertised Window parameter in the TCP segment carrying the acknowledge, but it does not affect the TCP protocol. The proposed technique is implemented in the network access gateways, that is, in the gateways through which both the incoming and outgoing packets related to a given TCP connection are forced to go, whatever the routing strategy used in the network. We show that when the access gateways implementing AWM are the bottleneck in the networks, TCP performance is very close to that of a pseudo constant bit rate protocol providing no loss, while network utilization is close to one.

A Markov model of a freerider in a BitTorrent P2P network

BitTorrent is today one of the largest P2P systems which allows file sharing for Internet users. Very little effort has been dedicated to this target up to now. The goal of this paper is to develop an analytical model of a free-rider in a BitTorrent network. Unlike previous analytical models which capture the behavior of the network as a whole, the proposed model is able to analyze the performance from the user perspective. The model is applied to a case study to evaluate performance in a real case, and to obtain some insights into the influence of BitTorrent parameters on system performance.

CLAPS: A Cross-Layer Analysis Platform for P2P Video Streaming

Peer-to-peer (P2P) networks are emerging as an interesting alternative to media streaming delivery with respect to content delivery networks (CDN). Although several P2P protocols have been proposed to support video streaming with an adequate degree of quality of service (QoS), a critical point today is how to evaluate them and optimize their parameters. The difficulty lies in the fact that the behavior of the P2P overlay network and the one of the underlying packet network are closely related and need to be analyzed simultaneously. Ad-hoc simulators for each P2P architecture have been proposed but they neglect the interactions between peers and the underlying network. This paper proposes a new approach to P2P video streaming performance analysis and system design which provides a very accurate cross-layer simulation of the real behavior of P2P clients over TCP/IP networks. A fluid-flow approach has been adopted to simulate the behavior of the network elements supporting an emulated overlay network consisting of real implementations of P2P clients. The P2P SplitStream video streaming protocol is considered as a case study to demonstrate that the tool is able to capture performance parameters at both the overlay (e.g. inter description synchronization) and packet network levels (e.g. traffic source round-trip time and router queue behavior).

A fluid-based model of time-limited TCP flows

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.

Improving fairness in a WRED-based DiffServ network: A fluid-flow approach

The DiffServ architecture has been proposed as a scalable approach for upgrading the Internet, adding service differentiation functionalities. However, several aspects of this architecture still have to be analyzed and solved. For this reason, network designers need to be provided with tools which are able to estimate the average behavior of a DiffServ network with a high level of accuracy and in a short time. In this paper a fluid-flow model of a DiffServ network supporting Assured Forwarding Per-Hop Behavior (PHB) and loaded with TCP flows is proposed. At the edge of the network, two rate three color markers (TRTCM) are employed as profile meters, while within the network core routers implement a Weighted RED (WRED) buffer management mechanism. In order to demonstrate the high accuracy of the proposed model, a comparison between model and simulation results is performed, taking into account not just a bottleneck link, but a complex network topology. The proposed analytical framework is then used to analyze the impact of several factors on the fair sharing of network resources between traffic aggregates with the same traffic profile, and to achieve some guidelines for WRED parameter settings with the aim of reducing unfairness.

Active window management: performance assessment through an extensive comparison with XCP

The most efficient approaches defined so far to address performance degradations in end-to-end congestion control exploit the flow control mechanism to improve end-to-end performance. The most authoritative solution in this context seems to be the eXplicit Control Protocol (XCP) which achieves high performance but requires changes in both network routers and hosts which make it difficult to deploy. To this aim we have developed a new mechanism, called Active Window Management (AWM), which is able to maintain the queue length in network routers almost constant providing no loss, while maximizing network utilization. The idea at the basis of AWM is to allow network routers to manipulate the Advertised Window field in TCP ACKs. In this way no modifications to the TCP protocol are required. The target of this paper is to propose an extensive numerical analysis of AWM to compare it with the XCP protocol, chosen as reference case.

Queue stability analysis and performance evaluation of a TCP-compliant window management mechanism

Addressing performance degradations in end-to-end congestion control has been one of the most active research areas in the last decade. A huge number of techniques have been proposed in the past, but some of them miss the challenging target of both minimizing network delay and keeping goodput close to the network capacity (e.g., active queue management (AQM) techniques); others require changes in both network routers and hosts that make them difficult to deploy [e.g., eXplicit Control Protocol (XCP)]. In a previous paper, a novel mechanism, called Active Window Management (AWM), has been proposed by the same authors with the aim of controlling the queue length in network routers, maintaining it almost constant to achieve no loss, while maximizing network utilization. The target of this paper is twofold: 1) carrying out a deep analysis of the AWM algorithm stability to evaluate the impact of some system parameters on the performance of the whole system; 2) demonstrating that AWM can be gradually introduced, maintaining the compatibility with the current Internet. An extensive numerical analysis is proposed to demonstrate that AWM stabilizes the queue around a target value with small oscillations.

An Analytical Model of a BitTorrent Peer

In this paper we propose a Markovian model of BitTorrent. Unlike already developed works which capture demographic dynamics, it focuses on the behavior of individual peers. To this end, we center our attention on a generic peer, called tagged peer (TP); for each possible logical state of a BT peer-to-peer connection maintained by the TP, we consider a stochastic process which counts the number of such links, and characterize them according to their state. Validation is carried out and steady-state analysis is performed in order to illustrate how performance evaluation can be extracted from our model

A New Fluid-Based Methodology to Model AQM Techniques with Markov Arrivals

In the present Internet, due to the popularity of AQM techniques, great effort is being devoting to analyzing their performance and optimizing their parameters. In this perspective the target of the paper is to provide an analytical framework to evaluate the performance of an AQM router loaded by any kind of traffic which can be modeled with a Markov arrival process. In order for complexity of the model to be independent of the queue size, the model uses a fluid-flow approach. In the paper, just to provide an example of application, we will consider RED routers. After the model definition, with the aim of demonstrating the correctness of the proposed methodology, the case of a RED buffer loaded by a constant bit-rate (CBR) source has been studied. In this particular case, where the classical theory of probability can be applied, the solution of the problem is derived, verifying that obtained analytical equations coincide with those obtained in the general case.

Fluid flow analysis of TCP flows in a DiffServ environment

The differentiated services (DiffServ) architecture allows IP networks to offer different QoS levels to different users and applications. In this architecture, routers in the core network offer the same per-hop behaviour (PHB) to all packets classified as belonging to the same class at the edge of the network. One of the most important types of PHB is assured forwarding (AF) PHB. Within each AF class, IP packets can be marked with different drop precedence (DP) values, and treated differently in congested DS nodes. To this end, DiffServ nodes in the core network implement active queue management (AQM) mechanisms. The target of this paper is to provide network designers with an accurate fluid-flow analytical model of a DiffServ network, where the RED with in/out and coupled average queues (RIO-C), RED with in/out and decoupled average queues (RIO-DC) and Weighted RED (WRED) AQM techniques are implemented. We address a network simultaneously loaded with both greedy and data-limited TCP flows, and we consider one AF class in which two DPs are defined, one for packets complying with the negotiated profile (IN packets), and another for packets that do not respect it (OUT packets). A token bucket marking mechanism is modelled for this purpose. The proposed model is applied to a complex network topology. Comparison between model and simulation results demonstrates that the model is able to capture both transient and steady-state network behaviour with a high degree of accuracy, even when not all network routers implement the same AQM technique. These characteristics make our modelling approach suitable to address the issue of network parameter optimisation. As an example, the link capacity dimensioning problem in a DiffServ domain by means of an iterative optimisation algorithm is presented. Copyright