Skip to main content
main-content

Über dieses Buch

The 2010 edition of the Passive and Active Measurement Conference was the 11th of a series of successful events. Since 2000, the Passive and Active M- surement (PAM) conference has provided a forum for presenting and discussing innovative and early work in the area of Internet measurements. PAM has a tradition of being a workshop-like conference with lively discussion and active participation from all attendees. This event focuses on research and practical applications of network measurement and analysis techniques. This year’s c- ference was held at ETH Zurich, Switzerland. PAM 2010 attracted 79 submissions. Each paper was carefully reviewed by at leastthreemembersoftheTechnicalProgramCommittee.Thereviewingprocess ledtotheacceptanceof23papers.Thepaperswerearrangedinninesessionsc- ering the following areas: routing, transport protocols, mobile devices, topology, measurement infrastructure, characterizing network usage, analysis techniques, tra?c analysis, and the Web. We are very grateful to Endace Ltd. (New Zealand), Cisco Systems Inc. (USA), armasuisse (Switzerland) and the COST Action TMA whose sponsoring allowedustokeepregistrationcostslowandto o?erseveraltravelgrantsto PhD students. We are also grateful to ETH Zurich for sponsoring PAM as a host.

Inhaltsverzeichnis

Frontmatter

Characterizing the Global Impact of P2P Overlays on the AS-Level Underlay

This paper examines the problem of characterizing and assessing the global impact of the load imposed by a Peer-to-Peer (P2P) overlay on the AS-level underlay. In particular, we capture Gnutella snapshots for four consecutive years, obtain the corresponding AS-level topology snapshots of the Internet and infer the AS-paths associated with each overlay connection. Assuming a simple model of overlay traffic, we analyze the observed load imposed by these Gnutella snapshots on the AS-level underlay using metrics that characterize the load seen on individual AS-paths and by the transit ASes, illustrate the churn among the top transit ASes during this 4-year period, and describe the propagation of traffic within the AS-level hierarchy.

Amir Hassan Rasti, Reza Rejaie, Walter Willinger

Investigating Occurrence of Duplicate Updates in BGP Announcements

BGP is a hard-state protocol that uses TCP connections to reliably exchange routing state updates between neighbor BGP routers. According to the protocol, only routing changes should trigger a BGP router to generate updates; updates that do not express any routing changes are superfluous and should not occur. Nonetheless, such ‘duplicate’ BGP updates have been observed in reports as early as 1998 and as recently as 2007. To date, no quantitative measurement has been conducted on how many of these duplicates get sent, who is sending them, when they are observed, what impact they have on the global health of the Internet, or why these ‘duplicate’ updates are even being generated. In this paper, we address all of the above through a systematic assessment on the BGP duplicate updates. We first show that duplicates can have a negative impact on router processing loads; routers can receive upto 86.42% duplicates during their busiest times. We then reveal that there is a significant number of duplicates on the Internet - about 13% of all BGP routing updates are duplicates. Finally, through a detailed investigation of duplicate properties, we manage to discover the major cause behind the generation of pathological duplicate BGP updates.

Jong Han Park, Dan Jen, Mohit Lad, Shane Amante, Danny McPherson, Lixia Zhang

A Measurement Study of the Origins of End-to-End Delay Variations

The end-to-end (e2e) stability of Internet routing has been studied for over a decade, focusing on routes and delays. This paper presents a novel technique for uncovering the origins of delay variations by measuring the overlap between delay distribution of probed routes, and how these are affected by route stability.

Evaluation is performed using two large scale experiments from 2006 and 2009, each measuring between more than 100 broadly distributed vantage points. Our main finding is that in both years, about 70% of the measured source-destination pairs and roughly 95% of the academic pairs, have delay variations mostly within the routes, while only 15-20% of the pairs and less than 5% of the academic pairs witness a clear difference between the delays of different routes.

Yaron Schwartz, Yuval Shavitt, Udi Weinsberg

Yes, We LEDBAT: Playing with the New BitTorrent Congestion Control Algorithm

Since December 2008, the official BitTorrent client is using a new congestion-control protocol for data transfer, implemented at the application layer and built over UDP at the transport-layer: this new protocol undergoes the name of LEDBAT, for Low Extra Delay Background Transport.

In this paper, we study different flavors of the LEDBAT protocol, corresponding to different milestones in the BitTorrent software evolution, by means of an active testbed. Focusing on single flow scenario, we investigate emulated artificial network conditions, such as additional delay and capacity limitation. Then, in order to better grasp the potential impact of LEDBAT on the current Internet traffic, we consider a multiple flows scenario, and investigate the performance of a mixture of TCP and LEDBAT flows, so to better assess what “lower-than best effort” means in practice. Our results show that LEDBAT has already fulfilled some of its original design goals, though some issues still need to be addressed.

Dario Rossi, Claudio Testa, Silvio Valenti

Measuring and Evaluating TCP Splitting for Cloud Services

In this paper, we examine the benefits of split-TCP proxies, deployed in an operational world-wide network, for accelerating cloud services. We consider a fraction of a network consisting of a large number of

satellite

datacenters, which host split-TCP proxies, and a smaller number of

mega

datacenters, which ultimately perform computation or provide storage. Using web search as an exemplary case study, our detailed measurements reveal that a vanilla TCP splitting solution deployed at the satellite DCs reduces the 95

th

percentile of latency by as much as 43% when compared to serving queries directly from the mega DCs. Through careful dissection of the measurement results, we characterize how individual components, including proxy stacks, network protocols, packet losses and network load, can impact the latency. Finally, we shed light on further optimizations that can fully realize the potential of the TCP splitting solution.

Abhinav Pathak, Y. Angela Wang, Cheng Huang, Albert Greenberg, Y. Charlie Hu, Randy Kern, Jin Li, Keith W. Ross

The Myth of Spatial Reuse with Directional Antennas in Indoor Wireless Networks

Interference among co-channel users is a fundamental problem in wireless networks, which prevents nearby links from operating concurrently. Directional antennas allow the radiation patterns of wireless transmitters to be shaped to form directed beams. Conventionally, such beams are assumed to improve the spatial reuse (i.e. concurrency) in indoor wireless networks. In this paper, we use experiments in an indoor office setting of Wifi Access points equipped with directional antennas, to study their potential for interference mitigation and spatial reuse. In contrast to conventional wisdom, we observe that the interference mitigation benefits of directional antennas are minimal. On analyzing our experimental traces we observe that directional links do

not

reduce interference to nearby links due to the lack of signal confinement due to indoor multipath fading. We then use the insights derived from our study to develop an alternative approach that provides better interference reduction in indoor networks compared to directional links.

Sriram Lakshmanan, Karthikeyan Sundaresan, Sampath Rangarajan, Raghupathy Sivakumar

Influence of the Packet Size on the One-Way Delay in 3G Networks

We currently observe a rising interest in mobile broadband, which users expect to perform in a similar way as its fixed counterpart. On the other hand, the capacity allocation process on mobile access links is far less transparent to the user; still, its properties need to be known in order to minimize the impact of the network on application performance. This paper investigates the impact of the packet size on the minimal one-way delay for the uplink in third-generation mobile networks. For interactive and real-time applications such as VoIP, one-way delays are of major importance for user perception; however, they are challenging to measure due to their sensitivity to clock synchronisation. Therefore, the paper applies a robust and innovative method to assure the quality of these measurements. Results from measurements from several Swedish mobile operators show that applications can gain significantly in terms of one-way delay from choosing optimal packet sizes. We show that, in certain cases, an increased packet size can improve the one-way delay performance at best by several hundred milliseconds.

Patrik Arlos, Markus Fiedler

An Experimental Performance Comparison of 3G and Wi-Fi

Mobile Internet users have two options for connectivity: pay premium fees to utilize 3G or wander around looking for open Wi-Fi access points. We perform an experimental evaluation of the amount of data that can be pushed to and pulled from the Internet on 3G and open Wi-Fi access points while on the move. This side-by-side comparison is carried out at both driving and walking speeds in an urban area using standard devices. We show that significant amounts of data can be transferred opportunistically without the need of always being connected to the network. We also show that Wi-Fi mostly suffers from not being able to exploit short contacts with access points but performs comparably well against 3G when downloading and even significantly better while uploading data.

Richard Gass, Christophe Diot

Extracting Intra-domain Topology from mrinfo Probing

Active and passive measurements for topology discovery have known an impressive growth during the last decade. If a lot of work has been done regarding inter-domain topology discovery and modeling, only a few papers raise the question of how to extract intra-domain topologies from measurements results.

In this paper, based on a large dataset collected with

mrinfo

, a multicast tool that silently discovers all interfaces of a router, we provide a mechanism for retrieving intra-domain topologies. The main challenge is to assign an AS number to a border router whose IP addresses are not mapped to the same AS. Our algorithm is based on probabilistic and empirical IP allocation rules. The goal of our pool of rules is to converge to a consistent router to AS mapping. We show that our router-to-AS algorithm results in a mapping in more than 99% of the cases. Furthermore, with

mrinfo

, point-to-point links between routers can be distinguished from multiple links attached to a switch, providing an accurate view of the collected topologies. Finally, we provide a set of large intra-domain topologies in various formats.

Jean-Jacques Pansiot, Pascal Mérindol, Benoit Donnet, Olivier Bonaventure

Quantifying the Pitfalls of Traceroute in AS Connectivity Inference

Although traceroute has the potential to discover AS links that are invisible to existing BGP monitors, it is well known that the common approach for mapping router IP address to AS number (IP2AS) based on the

longest prefix matching

is highly error-prone. In this paper we conduct a systematic investigation into the potential errors of the IP2AS mapping for AS topology inference. In comparing traceroute-derived AS paths and BGP AS paths, we take a novel approach of identifying mismatch fragments between each path pair. We then identify the origin and cause of each mismatch with a systematic set of tests based on publicly available data sets. Our results show that about 60% of mismatches are due to IP address sharing between peering BGP routers in neighboring ASes, and only about 14% of the mismatches are caused by the presence of IXPs, siblings, or prefixes with multiple origin ASes. This result helps clarify an argument that comes from previous work regarding the major cause of errors in converting traceroute paths to AS paths. Our results also show that between 16% and 47% of AS adjacencies in two public repositories for traceroute-derived topology are false.

Yu Zhang, Ricardo Oliveira, Hongli Zhang, Lixia Zhang

Toward Topology Dualism: Improving the Accuracy of AS Annotations for Routers

To describe, analyze, and model the topological and structural characteristics of the Internet, researchers use Internet maps constructed at the router or autonomous system (AS) level. Although progress has been made on each front individually, a

dual graph

representing connectivity of routers with AS labels remains an elusive goal. We take steps toward merging the router-level and AS-level views of the Internet. We start from a collection of traces, i.e. sequences of IP addresses obtained with large-scale traceroute measurements from a distributed set of vantage points. We use state-of-the-art alias resolution techniques to identify interfaces belonging to the same router. We develop novel heuristics to assign routers to ASes, producing an

AS-router dual graph

. We validate our router assignment heuristics using data provided by tier-1 and tier-2 ISPs and five research networks, and show that we successfully assign 80% of routers with interfaces from multiple ASes to the correct AS. When we include routers with interfaces from a single AS, the accuracy drops to 71%, due to the 24% of total inferred routers for which our measurement or alias resolution fails to find an interface belonging to the correct AS. We use our dual graph construct to estimate economic properties of the AS-router dual graph, such as the number of internal and border routers owned by different types of ASes. We also demonstrate how our techniques can improve IP-AS mapping, including resolving up to 62% of false loops we observed in AS paths derived from traceroutes.

Bradley Huffaker, Amogh Dhamdhere, Marina Fomenkov, kc claffy

The RIPE NCC Internet Measurement Data Repository

This paper describes datasets that will shortly be made available to the research community through an Internet measurement data repository operated by the RIPE NCC. The datasets include measurements collected by RIPE NCC projects, packet trace sets recovered from the defunct NLANR website and datasets collected and currently hosted by other research institutions. This work aims to raise awareness of these datasets amongst researchers and to promote discussion about possible changes to the data collection processes to ensure that the measurements are relevant and useful to the community.

Tony McGregor, Shane Alcock, Daniel Karrenberg

Enabling High-Performance Internet-Wide Measurements on Windows

This paper presents analysis of the Windows kernel network stack and designs a novel high-performance NDIS driver platform called IRLstack whose goal is to enable large-scale Internet measurements that require sending billions of packets and managing millions of outstanding connections on inexpensive commodity hardware available to any research lab. Our results show that with just 75% of one modern CPU core, IRLstack can saturate a gigabit link with SYN packets (i.e., 1.48M pps) and achieve 3.52 Gbps (i.e., 5.25 Mpps) with a quad-core CPU. IRLstack’s transmission performance exceeds that of Winsock by a factor of 92-174, batch-mode WinPcap by a factor of 4.7-6.7, and the latest optimized PF_RING/TNAPI Linux kernel by up to 30%.

Matt Smith, Dmitri Loguinov

MOR: Monitoring and Measurements through the Onion Router

A free and easy to use distributed monitoring and measurement platform would be valuable in several applications: monitoring network or server infrastructures, performing research experiments using many ISPs and test nodes, or checking for network neutrality violations performed by service providers. In this paper we present MOR, a technique for performing distributed measurement and monitoring tasks using the geographically diverse infrastructure of the Tor anonymizing network. Through several case studies, we show the applicability and value of MOR in revealing the structure and function of large hosting infrastructures and detecting network neutrality violations. Our experiments show that about 7.5% of the tested organizations block at least one popular application port and about 5.5% of them modify HTTP headers.

Demetris Antoniades, Evangelos P. Markatos, Constantine Dovrolis

Evaluating IPv6 Adoption in the Internet

As IPv4 address space approaches exhaustion, large networks are deploying IPv6 or preparing for deployment. However, there is little data available about the quantity and quality of IPv6 connectivity. We describe a methodology to measure IPv6 adoption from the perspective of a Web site operator and to evaluate the impact that adding IPv6 to a Web site will have on its users. We apply our methodology to the Google Web site and present results collected over the last year. Our data show that IPv6 adoption, while growing significantly, is still low, varies considerably by country, and is heavily influenced by a small number of large deployments. We find that native IPv6 latency is comparable to IPv4 and provide statistics on IPv6 transition mechanisms used.

Lorenzo Colitti, Steinar H. Gunderson, Erik Kline, Tiziana Refice

Internet Usage at Elementary, Middle and High Schools: A First Look at K-12 Traffic from Two US Georgia Counties

Earlier Internet traffic analysis studies have focused on enterprises [1,6], backbone networks [2,3], universities [5,7], or residential traffic [4]. However, much less is known about Internet usage in the K-12 educational system (elementary, middle and high schools). In this paper, we present a first analysis of network traffic captured at two K-12 districts in the US state of Georgia, also comparing with similar traces collected at our university (Georgia Tech). An interesting point is that one of the two K-12 counties has limited Internet access capacity and it is congested during most of the workday. Further, both K-12 networks are heavily firewalled, using both port-based and content-based filters. The paper focuses on the host activity, utilization trends, user activity, application mix, flow characteristics and communication dispersion in these two K-12 networks.

Robert Miller, Warren Matthews, Constantine Dovrolis

A First Look at Mobile Hand-Held Device Traffic

Although mobile hand-held devices (MHDs) are ubiquitous today, little is know about how they are used—especially at home. In this paper, we cast a first look on mobile hand-held device usage from a network perspective. We base our study on anonymized packet level data representing more than 20,000 residential DSL customers. Our characterization of the traffic shows that MHDs are active on up to 3% of the monitored DSL lines. Mobile devices from Apple (i.e., iPhones and iPods) are, by a huge margin, the most commonly used MHDs and account for most of the traffic. We find that MHD traffic is dominated by multimedia content and downloads of mobile applications.

Gregor Maier, Fabian Schneider, Anja Feldmann

A Learning-Based Approach for IP Geolocation

The ability to pinpoint the geographic location of IP hosts is compelling for applications such as on-line advertising and network attack diagnosis. While prior methods can accurately identify the location of hosts in some regions of the Internet, they produce erroneous results when the delay or topology measurement on which they are based is limited. The hypothesis of our work is that the accuracy of IP geolocation can be improved through the creation of a flexible analytic framework that accommodates different types of geolocation information. In this paper, we describe a new framework for IP geolocation that reduces to a machine-learning classification problem. Our methodology considers a set of lightweight measurements from a set of known monitors to a target, and then classifies the location of that target based on the most probable geographic region given probability densities learned from a training set. For this study, we employ a Naive Bayes framework that has low computational complexity and enables additional environmental information to be easily added to enhance the classification process. To demonstrate the feasibility and accuracy of our approach, we test IP geolocation on over 16,000 routers given ping measurements from 78 monitors with known geographic placement. Our results show that the simple application of our method improves geolocation accuracy for over 96% of the nodes identified in our data set, with on average accuracy 70 miles closer to the true geographic location versus prior constraint-based geolocation. These results highlight the promise of our method and indicate how future expansion of the classifier can lead to further improvements in geolocation accuracy.

Brian Eriksson, Paul Barford, Joel Sommers, Robert Nowak

A Probabilistic Population Study of the Conficker-C Botnet

We estimate the number of active machines per hour infected with the Conficker-C worm, using a probability model of Conficker-C’s UDP P2P scanning behavior. For an observer with access to a proportion

δ

of monitored IPv4 space, we derive the distribution of the number of times a single infected host is observed scanning the monitored space, based on a study of the P2P protocol, and on network and behavioral variability by relative hour of the day. We use these distributional results in conjunction with the Lévy form of the Central Limit Theorem to estimate the total number of active hosts in a single hour. We apply the model to observed data from Conficker-C scans sent over a 51-day period (March 5th through April 24th, 2009) to a large private network.

Rhiannon Weaver

Network DVR: A Programmable Framework for Application-Aware Trace Collection

Network traces are essential for a wide range of network applications, including traffic analysis, network measurement, performance monitoring, and security analysis. Existing capture tools do not have sufficient built-in intelligence to understand these application requirements. Consequently, they are forced to collect

all

packet traces that might be useful at the finest granularity to meet a certain level of accuracy requirement. It is up to the network applications to process the per-flow traffic statistics and extract meaningful information. But for a number of applications, it is much more efficient to record packet sequences for flows that match some application-specific signatures, specified using for example regular expressions. A basic approach is to begin memory-copy (recording) when the first character of a regular expression is matched. However, often times, a matching eventually fails, thus consuming unnecessary memory resources during the interim. In this paper, we present a programmable

application-aware

triggered trace collection system called Network DVR that performs precisely the function of packet content recording based on user-specified trigger signatures. This in turn significantly reduces the number of memory copies that the system has to consume for valid trace collection, which has been shown previously as a key indicator of system performance [8]. We evaluated our Network DVR implementation on a practical application using 10 real datasets that were gathered from a large enterprise Internet gateway. In comparison to the basic approach in which the memory-copy starts immediately upon the first character match without

triggered-recording

, Network DVR was able to reduce the amount of memory-copies by a factor of over 500x on average across the 10 datasets and over 800x in the best case.

Chia-Wei Chang, Alexandre Gerber, Bill Lin, Subhabrata Sen, Oliver Spatscheck

OpenTM: Traffic Matrix Estimator for OpenFlow Networks

In this paper we present OpenTM, a traffic matrix estimation system for OpenFlow networks. OpenTM uses built-in features provided in OpenFlow switches to directly and accurately measure the traffic matrix with a low overhead. Additionally, OpenTM uses the routing information learned from the OpenFlow controller to intelligently choose the switches from which to obtain flow statistics, thus reducing the load on switching elements. We explore several algorithms for choosing which switches to query, and demonstrate that there is a trade-off between accuracy of measurements, and the worst case maximum load on individual switches,

i.e.

, the perfect load balancing scheme sometimes results in the worst estimate, and the best estimation can lead to worst case load distribution among switches. We show that a non-uniform distribution querying strategy that tends to query switches closer to the destination with a higher probability has a better performance compared to the uniform schemes. Our test-bed experiments show that for a stationary traffic matrix OpenTM normally converges within ten queries which is considerably faster than existing traffic matrix estimation techniques for traditional IP networks.

Amin Tootoonchian, Monia Ghobadi, Yashar Ganjali

Web Timeouts and Their Implications

Timeouts play a fundamental role in network protocols, controlling numerous aspects of host behavior at different layers of the protocol stack. Previous work has documented a class of Denial of Service (DoS) attacks that leverage timeouts to force a host to preserve state with a bare minimum level of interactivity with the attacker. This paper considers the vulnerability of operational Web servers to such attacks by comparing timeouts implemented in servers with the normal Web activity that informs our understanding as to the necessary length of timeouts. We then use these two results-which generally show that the timeouts in wide use are long relative to normal Web transactions-to devise a framework to augment static timeouts with both measurements of the system and particular policy decisions in times of high load.

Zakaria Al-Qudah, Michael Rabinovich, Mark Allman

A Longitudinal View of HTTP Traffic

In this paper we analyze three and a half years of HTTP traffic observed at a small research institute to characterize the evolution of various facets of web operation. While our dataset is modest in terms of user population, it is unique in its temporal breadth. We leverage the longitudinal data to study various characteristics of the traffic, from client and server behavior to object and connection characteristics. In addition, we assess how the delivery of content is structured across our datasets, including the use of browser caches, the efficacy of network-based proxy caches, and the use of content delivery networks. While each of the aspects we study has been investigated to some extent in prior work, our contribution is a unique long-term characterization.

Tom Callahan, Mark Allman, Vern Paxson

Backmatter

Weitere Informationen

Premium Partner

Neuer Inhalt

BranchenIndex Online

Die B2B-Firmensuche für Industrie und Wirtschaft: Kostenfrei in Firmenprofilen nach Lieferanten, Herstellern, Dienstleistern und Händlern recherchieren.

Whitepaper

- ANZEIGE -

Product Lifecycle Management im Konzernumfeld – Herausforderungen, Lösungsansätze und Handlungsempfehlungen

Für produzierende Unternehmen hat sich Product Lifecycle Management in den letzten Jahrzehnten in wachsendem Maße zu einem strategisch wichtigen Ansatz entwickelt. Forciert durch steigende Effektivitäts- und Effizienzanforderungen stellen viele Unternehmen ihre Product Lifecycle Management-Prozesse und -Informationssysteme auf den Prüfstand. Der vorliegende Beitrag beschreibt entlang eines etablierten Analyseframeworks Herausforderungen und Lösungsansätze im Product Lifecycle Management im Konzernumfeld.
Jetzt gratis downloaden!

Bildnachweise