David Ros

Dynamic Window Coupling for multipath congestion control

The traditional problem of end-hosts efficiently and fairly utilizing end-to-end paths becomes significantly harder when the end-hosts are multihomed. Such is the case, for instance, when an end-host has simultaneous connectivity through several service providers, or when a mobile device is simultaneously connected via both a wireless LAN and a cellular network. A multihoming-aware transport protocol, such as MPTCP or SCTP, that sends data over the multiple resulting end-to-end paths must be fair to other flows in the network while being able to maximize its own throughput. In this paper, we present Dynamic Window Coupling (DWC), a multipath congestion control mechanism that seeks to achieve both these goals. DWC uses loss and delay signals to detect shared bottlenecks, explicitly grouping and sharing congestion control across subflows on paths that have a common bottleneck, while separating congestion control for subflows on paths with distinct bottlenecks. DWC detects shifting bottlenecks in the network and responds by dynamically regrouping subflows. Simulations demonstrate that DWC detects shared bottlenecks under most network topologies and conditions that we considered, regroups subflows correctly as bottlenecks shift, aggregates throughput across distinct bottlenecks, and is fair to other TCP flows at all bottlenecks.

RSSI-Based Forwarding for Multihop Wireless Sensor Networks

In a multihop Wireless Sensor Network (WSN), a salient point among routing protocols that do not depend on network topology and existence of neighboring nodes is the need to know sensor nodes geographical location with respect to the sink node. This is obtained by some means like Global Positioning System (GPS) and localization techniques. In a prior work, we have proposed RSSI-based Forwarding (RBF) protocol that works without knowledge of nodes location by using a Received Signal Strength Indicator (RSSI) level of beacon signals transmitted by the sink. Through contention, a next-hop node is determined among the forwarding candidates using a timer-based suppression scheme. We propose an improvement of the suppression scheme in which a contender closer to the sink is favored with a higher probability for being selected as a next-hop node. By means of simulation, it is shown that the performance of RBF is significantly improved using the enhanced mechanism.

The New AQM Kids on the Block: An Experimental Evaluation of CoDel and PIE

Active Queue Management (AQM) design has again come into the spotlight of network operators, vendors and OS developers. This reflects the growing concern and sensitivity about the end-to-end latency perceived by todays Internet users. CoDel and PIE are two AQM mechanisms that have recently been presented and discussed in the IRTF and the IETF as solutions for keeping latency low. To the best of our knowledge, they have so far only been evaluated or compared against each other using default parameter settings, which naturally presents a rather limited view of their operational range. We set thus to perform a broader experimental evaluation using real-world implementations in a wired testbed. We have in addition compared them with a decade-old variant of RED called Adaptive RED, which shares with CoDel and PIE the goal of “knob-free” operation. Surprisingly, in several instances results were favorable towards Adaptive RED.

On the Impact of Random Losses on TCP Performance in Coded Wireless Mesh Networks

Network coding (NC) is a promising technique to improve throughput in wireless mesh networks. However, some previous studies have found that the actual performance improvements offered by NC may be much lower than the gains predicted by theory. This is especially so when the bulk of the traffic carried by the mesh network is composed of TCP flows. By means of both mathematical modeling and ns-2 simulations, we explore some issues due to random packet loss in coded mesh networks. Our results illustrate how the use of network coding may induce synchronization between TCP flows; they also suggest that, under particular conditions of random packet loss, the aggregate throughput may actually be lower when NC is used.

Loss synchronization and router buffer sizing with high-speed versions of TCP

In this paper we are interested in looking at packet loss synchronization, and the related issue of buffer sizing, when the bulk of the traffic is composed of a moderate number of flows using high-speed versions of TCP. In particular, we wanted to explore whether increased levels of synchronization would be an outcome of using more aggressive versions of TCP over high-bandwidth paths. By means of ns-2 simulations, we have evaluated several TCP versions and a wide range of buffer sizes, as well as three different drop synchronization metrics. Our preliminary findings suggest that high-speed versions of TCP do yield higher levels of synchronization. However, in spite of a strong drop synchronization, such TCP versions can achieve both high goodput and link utilization, as long as enough buffering is provided.

TCP over Low-Power and Lossy Networks: Tuning the Segment Size to Minimize Energy Consumption

Low-power and Lossy Networks (LLNs), like wireless networks based upon the IEEE 802.15.4 standard, have strong energy constraints, and are moreover subject to frequent transmission errors, not only due to congestion but also to collisions and to radio channel conditions. This paper introduces an analytical model to compute the total energy consumption in an LLN due to the TCP protocol. The model allows us to highlight some tradeoffs as regards the choice of the TCP maximum segment size, of the Forward Error Correction (FEC) redundancy ratio, and of the number of link-layer retransmissions, in order to minimize the total energy consumption.

An Adaptive RIO (A-RIO) Queue Management Algorithm

In the context of the DiffServ architecture, active queue management (AQM) algorithms are used for the differentiated forwarding of packets. However, correctly setting the parameters of an AQM algorithm may prove difficult and error-prone. Besides, many studies have shown that the performance of AQM mechanisms is very sensitive to network conditions. In this paper we present an adaptive AQM algorithm, which we call Adaptive RIO (A-RIO), addressing both of these problems. Our simulation results show that A-RIO outperforms RIO in terms of stabilizing the queue occupation (and, hence, queuing delay), while maintaining a high throughput and a good protection of high-priority packets; A-RIOcould then be used for building controlled-delay, AF-based services. These results also provide some engineering rules that may be applied to improve the behaviour of the classical, non-adaptive RIO.

Improving distributed TCP caching for wireless sensor networks

Reliable transport protocols like TCP are in principle not well suited for wireless sensor networks (WSNs), due to both the characteristics of the network nodes (low computing power, strong energy constraints) and those of the typical applications running on such nodes (low data rates). Indeed, TCP has not been designed with energy efficiency in mind, and the end-to-end recovery of TCP makes it too energy-expensive for common WSN applications. Distributed TCP Caching (DTC), originally proposed by Dunkels et al., is a mechanism aimed at adapting TCP to wireless sensor networks. DTC improves TCP performance by caching TCP segments in intermediate nodes. However, DTC was designed for, and tested with, a TDMA-based MAC layer without automatic repeat request (ARQ). On the other hand, current WSN link-layer protocols like 802.15.4 are based on Carrier Sense Multiple Access with Collision Avoidance (CSMA-CA), and they readily support ARQ. In this paper, we propose and evaluate several modifications to DTC that result in an enhanced, more energy-efficient transport layer for IP-based WSNs. We explicitly consider the CSMA-CA nature of the link (i.e., the presence of frame collisions), as well as the use of link-layer reliability. Using the INETMANET framework of the OMNET++ simulator, we assess how simple modifications to the TCP caching mechanism and round-trip time estimator can have a significant impact on DTC performance, in terms of energy efficiency and file transfer delay.

Impact of WiMAX Network Asymmetry on TCP

The IEEE 802.16 is the standard for broadband wireless access. One promise of this technology (also called WiMAX) is to provide high-speed access to the Internet where the transmission control protocol (TCP) is the core transport protocol. In this paper, we study the impact of network asymmetry in WiMAX on TCP performance. In particular, we investigate the dependence of the aggregate throughput and goodput of TCP on factors such as frame duration, direction of flow, DL:UL ratio, modulation and coding schemes, and offered loads. We find that these factors affect TCP performance by exacerbating the network asymmetry inherent to the MAC layer.

A simulation study of the Adaptive RIO (A-RIO) queue management algorithm

Active queue management (AQM) algorithms are useful not only for congestion avoidance purposes, but also for the differentiated forwarding of packets, as is done in the DiffServ architecture. It is well known that correctly setting the parameters of an AQM algorithm may prove difficult and error-prone. Besides, many studies have shown that the performance of AQM mechanisms is very sensitive to network conditions. In this paper we present a detailed simulation study of an Adaptive RIO (A-RIO) AQM algorithm which addresses both of these problems. A-RIO, first introduced by Orozco and Ros (2003), draws directly from the original RIO proposal of Clark and Fang (1998) and the Adaptive RED (A-RED) algorithm described by Floyd et al. (2001). Our results, based on ns-2 simulations, illustrate how A-RIO improves over RIO in terms of stabilizing the queue occupation (and, hence, queuing delay), while maintaining a high throughput and a good protection of high-priority packets; A-RIO could then be used for building controlled-delay, AF-based services. These results also provide some engineering rules that may be applied to improve the behaviour of the classical, non-adaptive RIO.