Renata Teixeira

Dynamics of hot-potato routing in IP networks

Despite the architectural separation between intradomain and interdomain routing in the Internet, intradomain protocols do influence the path-selection process in the Border Gateway Protocol (BGP). When choosing between multiple equally-good BGP routes, a router selects the one with the closest egress point, based on the intradomain path cost. Under such hot-potato routing, an intradomain event can trigger BGP routing changes. To characterize the influence of hot-potato routing, we conduct controlled experiments with a commercial router. Then, we propose a technique for associating BGP routing changes with events visible in the intradomain protocol, and apply our algorithm to AT&Ts backbone network. We show that (i) hot-potato routing can be a significant source of BGP updates, (ii) BGP updates can lag 60 seconds or more behind the intradomain event, (iii) the number of BGP path changes triggered by hot-potato routing has a nearly uniform distribution across destination prefixes, and (iv) the fraction of BGP messages triggered by intradomain changes varies significantly across time and router locations. We show that hot-potato routing changes lead to longer delays in forwarding-plane convergence, shifts in the flow of traffic to neighboring domains, extra externally-visible BGP update messages, and inaccuracies in Internet performance measurements.

Network sensitivity to hot-potato disruptions

Hot-potato routing is a mechanism employed when there are multiple (equally good) interdomain routes available for a given destination. In this scenario, the Border Gateway Protocol (BGP) selects the interdomain route associated with the closest egress point based upon intradomain path costs. Consequently, intradomain routing changes can impact interdomain routing and cause abrupt swings of external routes, which we call hot-potato disruptions. Recent work has shown that hot-potato disruptions can have a substantial impact on large ISP backbones and thereby jeopardize the network robustness. As a result, there is a need for guidelines and tools to assist in the design of networks that minimize hot-potato disruptions. However, developing these tools is challenging due to the complex and subtle nature of the interactions between exterior and interior routing. In this paper, we address these challenges using an analytic model of hot-potato routing that incorporates metrics to evaluate network sensitivity to hot-potato disruptions. We then present a methodology for computing these metrics using measurements of real ISP networks. We demonstrate the utility of our model by analyzing the sensitivity of a large AS in a tier~1 ISP network.

Characterizing and measuring path diversity of internet topologies

The use of multiple network paths between a pair of hosts has been proposed for a wide variety of network technologies in order to achieve higher bandwidth in data transfers, to select paths with low latency, to balance load in the network, and to protect against failures. Path diversity describes the number of disjoint paths between hosts and characteristics of those paths, such as path length, latency, and bottleneck bandwidth. Despite wide interest in such techniques, little work has been done that actually characterizes and measures the path diversity that exists in large-scale deployed packet networks like the Internet. Our work characterizes layer-3 path diversity 1 in the Internet.

Measuring multipath routing in the internet

Tools to measure Internet properties usually assume the existence of just one single path from a source to a destination. However, load-balancing capabilities, which create multiple active paths between two end-hosts, are available in most contemporary routers. This paper extends Paris trace route and proposes an extensive characterization of multipath routing in the Internet. We use Paris traceroute from RON and PlanetLab nodes to collect various datasets in 2007 and 2009. Our results show that the traditional concept of a single network path between hosts no longer holds. For instance, 39% of the source-destination pairs in our 2007 traces traverse a load balancer. This fraction increases to 72% if we consider the paths between a source and a destination network. In 2009, we notice a consolidation of per-flow and per-destination techniques and confirm that per-packet load balancing is rare.

Impact of hot-potato routing changes in IP networks

Despite the architectural separation between intradomain and interdomain routing in the Internet, intradomain protocols do influence the path-selection process in the Border Gateway Protocol (BGP). When choosing between multiple equally-good BGP routes, a router selects the one with the closest egress point, based on the intradomain path cost. Under such hot-potato routing, an intradomain event can trigger BGP routing changes. To characterize the influence of hot-potato routing, we propose a technique for associating BGP routing changes with events visible in the intradomain protocol, and apply our algorithm to a tier-1 ISP backbone network. We show that (i) BGP updates can lag 60 seconds or more behind the intradomain event; (ii) the number of BGP path changes triggered by hot-potato routing has a nearly uniform distribution across destination prefixes; and (iii) the fraction of BGP messages triggered by intradomain changes varies significantly across time and router locations. We show that hot-potato routing changes lead to longer delays in forwarding-plane convergence, shifts in the flow of traffic to neighboring domains, extra externally-visible BGP update messages, and inaccuracies in Internet performance measurements.

Probe and pray: using UPnP for home network measurements

Network measurement practitioners increasingly focus their interest on understanding and debugging home networks. The Universal Plug and Play (UPnP) technology holds promise as a highly efficient way to collect and leverage measurement data and configuration settings available from UPnP-enabled devices found in home networks. Unfortunately, UPnP proves less available and reliable than one would hope. In this paper, we explore the usability of UPnP as a means to measure and characterize home networks. We use data from 120,000 homes, collected with the HomeNet Profiler and Netalyzr troubleshooting suites. Our results show that in the majority of homes we could not collect any UPnP data at all, and when we could, the results were frequently inaccurate or simply wrong. Whenever UPnP-supplied data proved accurate, however, we demonstrate that UPnP provides an array of useful measurement techniques for inferring home network traffic and losses, for identifying home gateway models with configuration or implementation issues, and for obtaining ground truth on access link capacity.

TIE breaking: tunable interdomain egress selection

In a large backbone network, the routers often have multiple egress points they could use to direct traffic toward an external destination. Todays routers select the closest egress point, based on the intradomain routing configuration, in a practice known as early-exit or hot-potato routing. In this paper, we argue that hot-potato routing is restrictive, disruptive, and convoluted and propose an alternative called TIE (Tunable Interdomain Egress selection). TIE is a flexible mechanism that allows routers to select the egress point for each destination prefix based on both the intradomain topology and the goals of the network administrators. In fact, TIE is designed from the start with optimization in mind, to satisfy diverse requirements for traffic engineering and network robustness. We present two example optimization problems that use integer-programming and multicommodity-flow techniques, respectively, to tune the TIE mechanism to satisfy networkwide objectives. Experiments with traffic, topology, and routing data from two backbone networks demonstrate that our solution is both simple (for the routers) and expressive (for the network administrators).

Multipath tracing with Paris traceroute

Traceroute is a tool to report the route packets lake between two internet hosts. However, with the deployment of load balancing, there is no longer a single route to a destination, hence classic traceroute systematically misses some of these paths. In this paper, we specify an adaptive, stochastic probing algorithm, called the Multipath detection algorithm, to report all paths towards a destination. We have deployed this algorithm, probing from a single source towards multiple destinations. In our results, we have found Instances of load balancing with as many as 16 interfaces per hop. The algorithm also allows us to count load balancing routers, identify their locations, and characterize them by type.

BGP routing changes: merging views from two ISPs

Large ISPs experience millions of BGP routing changes a day. In this paper, we discuss the impact of BGP routing changes on the flow of trafffic, summarizing and reconciling the results from six measurement studies of the Sprint and AT&T backbone networks.

Implementation Issues of Early Application Identification

The automatic identification of applications associated with network traffic is an essential step to apply quality-of-service policies and profile network usage. Our prior work proposes Early Application Identification , a method that accurately identifies the application after the first four packets of a TCP connection. However, an online implementation of this method faces two challenges: it needs to run at high speed and with limited memory. This paper addresses these issues. We propose an algorithm that implements Early Application Identification plus a number of computation and memory optimizations. An evaluation using traffic traces collected at our university network shows that this implementation can classify traffic at up to 6 Gbit/s. This speed is more than enough to classify traffic at current edge networks.