The contents of this legacy page are no longer maintained nor supported, and are made available only for historical purposes.

Research Community Use of skitter Datasets

Skitter datasets are made available to multiple researchers in the community via CAIDA's FTP site. If you are interested in obtaining skitter datasets for your research, please read the Acceptable Use Policy.

Current community use of skitter datasets are briefly summarized below.

  • AFIT, Computer Engineering
    Ben Musial wishes to use the data for network visualization research. This initial research is a term project which will be further developed in his thesis research later this year. His intial plans are to display the network with data from a few thousand nodes and use a variety of visualization techniques to explore the data. Details of implementation will be largely time frame dependent (I've only got a few weeks to built the initial tool). The proposal for the thesis hasn't been developed yet and will be dependent on any results obtained from this work.

  • Ajou University, Korea
    Kim Soondong is studying DDoS and needs routing infomation. Routing infomation PI marking scheme uses router's IP address. Spoofed IP address does not easy to trace back the source. So PI mark hashed values of each routers. I want to expriment PI scheme with different points of view. [December 2003]

    Lee Hyun Ju read a paper titled "pi: A path identification mechanism to Defend against DDoS Attacks". At the paper, the authors experiment using Internet Map and Skitter map. I'll also experiment using the Map. And I'll also write a paper using the experiment" My basic idea may be summarized as follows. I chose n=2 bit shceme. I'll make 4 by 8 matrix. Where, row is 4(00, 01, 10, 11) and column is 8(16 bit/2 bit). I'll monitor the pi marking. And I'll count each section's value.(initial value is 0) [October 2003]

  • Ben-Gurion University of the Negev
    Michael Segal: The explosive growth of the internet has been accompanied by a wide range of internetworking problems related to routing, resource reservation, and administration, all these creates new engineering challenges. Facing these challenges and seeking for new solutions often involves simulation or analysis based on abstractions or models of the actual network structure. The reason is clear: networks that are large enough to be relevant to these kind of studies are also expensive and difficult to control, therefore they are rarely available for experimental purposes. It is therefore remarkable that studies based on randomly generated or trivial network topologies are so common, while rigorous analyses of how the result scale or how they might change with different topology are so rare.

    Still, one of the most profound questions in science is how a collection of elements self-organize to form new and extremely complex systems such as networks. Thus, networks are not only interesting from the communication point of view as they represent today a common feature of interest in several branches of science. The study of networks is widespread in social sciences, mathematics, physics and biology. Most remarkably, during the last years this approach has increased across the biology sciences. It is now clear that biological networks span the widest range of scales, from large food webs to molecular webs of chemical reactions between enzymes and their substrates. Among all these networks, there is one system that its analogy to communication networks is impossible to omit, the neuronal system. In a very simplistic view, neuronal networks are composed of different types of cells hard-wired by means long protuberances named axons and dendrites. The system is able to process and store information thanks to its delicate and complex connectivity. But even though there is a large amount of biological knowledge, there is still an open question of how functioning neuronal networks -ranging from small circuits to the whole brain- emerge from a collection of single entities, the individual neurons. As in networks in general, there is a strong relation between the neuronal network structure, or wiring diagram, and its function, i.e. the form-function relation. A deep understanding of this relation might be the key to determine the activity dynamics and the processing capabilities of biological neuronal networks. Biological systems are known for their capability to perform complex task while minimizing energetic costs as a result of millions of years of evolution. In the other hand all the human made technological networks (including the internet) where developed just in the recent years. So by comparing both neuronal and internet networks we can improve our knowledge on the two systems and maybe help us understand some of the biological reasons that determine the structure of the neuronal network, and maybe even help us develop new rules for large internet network design. [December 2003]

  • BT Advanced Communications Centre
    Jim Rudkin is conducting research into the stability of Internet routing. IP being a stateless protocol, packets can take different paths through the network to the same endpoint. He is interested in whether this type of event happens frequently or very rarely in the real world. The frequency of skitter measurements should provide an indication of how often routing changes occur.

  • Boston University
    Karim Mattar and Azer Bestavros are working on formulating an analytical model for P2P networks and wish to model the core network as a component of the model. They need to determine the average degree of the router network to choose an appropriate mathematical model. This can be partially obtained by analysing the router map. [April 2004]

    Anukool Lakhina used skitter data to prepare a combined journal version of two related papers: On the Marginal Utility of Deploying Measurement Infrastructure (IMW'01) and Sampling Biases in IP Topology Measurements (Infocom 03). The first paper examines the utility derived from adding sources and destinations in a measurement infrastructure. The second paper studies the biases that may result when few sources and destinations are used when mapping the IP Topology. For the journal version, he used an iffinder-disambiguated skitter dataset. [June 2003]

    Dhiman Barman distributions of round trip delays of TCP or one way delays of internet paths for his research. [March 2003]

    Shudong Jin studies IP connectivity. [November 2002]

    Professors John Byers, Mark Crovella and Azer Bestavros together with Prof. Paul Barford at the University of Wisconsin have used the skitter dataset to study the marginal utility derived from additional network topology measurements. They have focused on the comparative effectiveness of strategies for mapping the backbone of the Internet topology from a set of distributed vantage points. Their current work seeks to understand Internet scaling laws from these datasets, i.e. how many distinct routers and links one can expect to "see" by a set of traceroutes conducted from distributed measurement sites. See [July 2001]

    Paul Barford, John Byers, Mark Crovella and Ibrahim Matta have studied the marginal utility of network topology measurements using data from the Skitter project. Their work has centered on the manner in which additional nodes and edges of the Internet router graph are exposed by adding sources and destinations to the Skitter topology. Their results show that the utility of additional sources drops off considerably after the first two to three sources. Their results also expose the distinction between the Internet's "switching core" and its feeder networks, with declining marginal utility of measurements particularly true for switching core nodes and edges. Skitter data are the principal inputs to this study; furthermore, no equivalent source of data capable of supporting these questions is available.

  • Brno University of Technology, Brno, Czechoslovakia
    Petr Fiedler uses raw skitter data to verify theories presented in my PhD thesis. His work concerns improving of clock synchronization over non-deterministic networks with varying delays based on statistical properties of the round trip delays. [May 2003]

  • Carnegie Mellon University
    Kunwadee Sripanidkulch is working on designing overlay networks. The goal is to identify sound design principles based on data obtained from real overlay deployments, come up with new designs, and evaluate them in simulation. The skitter data can provide realistic topologies that can be used in the evaluation. [February 2004]

    Cynthia Wong is investigating the effects of worm spread by immunization on different nodes. I'd like to use this data for running simulations. [November 2003]

    Geoffrey Xie, visiting from the Naval Postgraduate School, wants to evaluate the accuracy of the existing models for predicting global BGP routing instabilities that will severely degrade the performance of user traffic. The historic data collected by skitter provides good samples to test several hypotheses. [October 2003]
    XiaFeng Wang and Prof. Mike reiter at ECE department, Carnegie Mellon University wish to use skitter data to study the countermeasures to distributed denial-of-service attacks. We want to evaluate our approaches in a realistic Internet topology. [ Abraham Yaar studies PI (short for Path Identifier) is a network infrastructure mechanism for mitigating DDoS attacks. Pi-enabled routers mark all packets they forward with a number of deterministic bits. When markings from multiple routers are aggregated, each packet has a "fingerprint" (or Pi-mark) based on the path it took through the Internet. Once attack packets are identified by the DDoS victim, subsequent attack packets can easily be filtered by the victim or its ISP, simply by matching the Pi-mark to the known attack packets. The Skitter datasets are important to our work because they represent a subset of real Internet topology, which we will use to test our implementation of Pi. [July 2003]

    Michael Collins of CERT and CMU is working with Dr. Michael Reiter to validate various forms of DOS detection/filtering algorithms. Using the skitter data, he plans to build a small map of addresses seen in DDOS attacks to understand their distribution across the network. [May 2003]

    Adrian Perrig uses Skitter data to help us better understand Internet Topology, specifically as it applies to IP Traceback and its performance. The use of "real world" topology is critical to a proper analysis of current techniques as well as in developing future models of Distributed Denial of Service Attacks. [January 2003]

    Dawn Song uses skitter data to new IP traceback schemes and defense schemes against DDoS attacks. [January 2003]

    Tze Sing Eugene Ng plans to use the skitter data to better understand the global network performance from the perspectives of different observation points. Strategy involves obtaining statistics that may characterize such global performance, and then making use of these statistics to guide algorithm designs for applications that span the global network.

    Yang-hua Chu and Sanjay Rao (with advisor Professor Hui Zhang) are conducting a simulation study of multicast routing using the public skitter dataset for "realistic" topology in the simulator, see

    Aditya Akella and Srinivasan Seshan are idenfitying peering points in the Internet and characterizing how losses are distributed. [June 2002]

    Giuseppe Di Fatta, Giuseppe Lo Re and Giuseppe Carollo at CERE-CNR in Palermo Italy note that their interest in Internet topology discovery raises from several factors:

    1) maintaining a database of the topology and tracing its evolution;
    2) understanding topological properties of real networks;
    3) collecting real topologies to be used to test protocols and network algorithm;

    They are carrying out experiments and distributed traceroute probes on the Internet. Skitter data that will be used for comparison purpose and to validate their approach. For the CERE-CNR "Topology Discovery" project, see

  • Chinese Academy of Sciences Institute of Software
    Dequan Li is working on a IP Traceback algorithm for DDoS attack tracing. He needs Internet topology data to validate his method. [November 2002]

  • The Chinese University of Hong Kong
    Ken Hui use data for massive graph research. In our research, we would like to start collecting different current network topology to investigate the current network model, e.g scale free model or small world model. Then we will try to summarize the topology usage and current suitation. After that, we would like to propose algorithm on current network topology. [March 2003]

    Ma Tianbai uses skitter data to analyze small world phenomena in a large scale network as a project for a course Massive graph and networking. [March 2003]

    Wendy Garvin Some Internet security incidents require triage notification to core Internet providers. Cisco plans to use the benchmarks created for Skitter to determine who is the "core" of the Internet and work a notification priority from the inside out - based on Skitter metrics. [October 2002]

    Barry Raveendran Greene (
    There are some Internet security incidents that require a triage notification of the core of the Internet. Cisco plans on using the benchmarks created for Skitter to determine who is the "core" of the Internet and work a notification priority from the inside out - base on Skitter metrics.

    Preliminary data based on the 2000 sample data was used for the recent SNMP Vulnerability. Cisco used Skitter as a reference point in it's converations with CERT and the US Government - demonstrating that a quantitative approachcan be used to determine the "core" (vs the notification on a whim method the US Government was taking).

  • Columbia University, NY
    Yechiam Yemini with the assistance of his student, Tsung-Liang Lu, will use the data to develop graph noise metrics and to characterize the Signal-Noise Ratio of the internet. After development of the graph noise metrics, we would like to incorporate and study the prefix structue to guide clustering; and, to study simplified internet routing. [June 2004]

  • CoRiTel, Italy
    Ivano Bartoli wants to better understand the characteristics of Internet paths in order to modeling the Internet in a realistic way. Further, we want to use this model to test the performance of real-time applicastions (such as videotelefony) over the Internet. [October 2003]

  • Delft University of Technology, The Netherlands
    Xiaoming Zhou, of the Network Architectures and Systems Department, performs statistical analysis of Internet Data. To enlarge the insight in how the global network behaves and evolves, he is going to use the Skitter data to statistically analyze the hopcount and end-to-end delay of the network between different locations around the world. For the "Statistical Analysis of Internet Data (SAID)" project, see [April 2003]

  • De Montfort University Communication Network Research Group, Software Technology Research Laboratory
    Doris Ressmann researches "reconfiguration of large, global distributed systems". She wrote a simulation to validate her theory, however this simulation needs some real life data. She will use the data, to simulate the delay of the network between different locations around the world.

    Doris will only be able to run her simulation on a campus wide network, but it has to react as when it would be distributed all over the world. With the time she plans to make some more information available on the following page: (link expired)

  • DoD Laboratory for Telecommunications Sciences
    Dr. William Semancik is the director of the Laboratory for Telecommunications Sciences, a laboratory performing network and telecommunications research on behalf of the Department of Defense. His group is running a project on the use of network topology as a means of creating an active cyber defense mechanism. This work is being performed both in-house and with Dr. Steve Kent of BBN Technologies. [January 2003]

  • Dresden University of Technology
    Frank Breiter's PhD Topic concerns analysis of new algorithms for QoS-routing over the Internet. Part of his work involves deriving the following statistics from skitter Internet topology data:

    1) mean number of paths between any source destination pair
    2) mean number of shortest path hops
    3) mean number of shortest paths between any source destination pair

  • EPFL Lausanne, France
    Hung Nguyen uses skitter data to test algorithms that find an optimal set of probes and an optimal set of beacons (measurment points) for multiple link failure diagnosis in a given topology. We have tested our algorithms on simulated networks and the tests showed good results. [December 2003]

  • ETHZ Swiss Federal Institute of Technology, Zurich, Switzerland
    Roger Wattenhofer analyzed Internet Topology: How does the Internet Topology look like when considering policies? Can the ASes be classified into several Tiers? What is the longest versus shortest backup path ? [September 2003]

    Vukadinovic Danica studiesthe number of disjoint paths and minimum vertex-cut between any pair of autonomous systems in the Internet when BGP policies are taken into account. Previously, he used AS graphs obtained from BGP tables, but we would like to use also traceroute AS graphs as a basis to infer AS relationships and compare our algorithm and analysis on both kind of graphs. [May 2003]

  • Federal University of Rio de Janeiro, Brazil
    Denilson Martins is researching IP traceback methods. [June 2003]

  • George Mason University
    Sean Gorman is working on an infrastructure mapping project at George Mason assessing Internet infrastructure. Two of the projects have used simulated IP topologies to (1) asses the cost effectiveness of various cyber defense strategies, and (2) measure the impact of diversity on security. It would be very valuable to run the simulations with CAIDA's skitter data instead of our mock data to both compare results and ascertain more realistic outcomes.

    A second less developed project examines the physical geography of interconnection. The method would look to ascertain the location of nodes when traffic passes from one provider to another. The hope is to develop a rough spatial topology map, albeit with only a sample of data for which we are able to determine a location. Papers for the first two projects with greater detail on the research are available if needed. [February 2004]

    Andy Liu uses skitter data for network visualization in intrusion detection research. His intial goal is to see the trend of intrusions on the network with data from a few thousand nodes and use a variety of visualization techniques to explore the data. Details of implementation will be largely time frame dependent.The proposal for the thesis hasn't been developed yet and will be dependent on any results obtained from this work. [October 2003]

    Xiaoxia Liu is investigating the impact of the Internet and information technology on regional economies. [November 2002]

  • Georgia Tech
    Kun Zhang and Jiang Xiao are going to use the skitter data to study the exchange points in the Internet. They expect to use the data to build intrusion detection system. [March 2003]

    Minho Sung and others are conducting research into a novel packet logging based (i.e., hash-based) traceback scheme that requires an order of magnitude smaller processing and storage cost than existing hash-based techniques. This new traceback scheme is likely to be scalable to much higher link speed (e.g., OC-768). We wish to use the skitter data to show in simulations that our scheme is effective and scalable. [November 2002]

    Ruomei Gao and Ellen Zegura are going to use the skitter data to study the exchange points in the Internet. They expect to use the data to discover the connectivity (topological charactors) of exchange points, and they also would like to explore other aspects of exchange points like traffic volumn, distribution, etc.

  • Huazhong University of Science and Technology, Wuhan City, Hubei Province, P.R. China
    Wu Weidong wants to investigate the BGP tables before and after using policy atoms. He wants to know how to group the prefixes after using policy atoms. [November 2002]

  • Institut Eurecom, Sophia-Antipolis, France
    Luis Garces-Erice designed a topology-aware DHT named TOPLUS, using IP prefixes from BGP tables, and plans to build an application level Multicast system on top of it. In order to perform realistic and fine-grained performance measurements of our system in an hypothetic Internet deployment, he plans to use the information gathered by Skitter. This will help prove the efficiency of the algorithms for locality-aware P2P network construction. [March 2004]

  • Institute of Computing Technology, P.R. China
    Jing Huang investigates whether Internet metrics such as RTT, route skip number, and change relate to each other, and whether they auto-correlate over time to some extent. If we make these relation clear, Internet will not look like so complex and unpredictable. [March 2003]

  • Indian Institute of Technology, Bombay, India,br /> Pavan Kumar would like to see if the skitter data will be useful for his thesis on network visualization research and traffic engineering [June 2004]

  • Indian Institute of Technology, Guwahati, India
    Shravan Kumar Rayanchu: The DDos attacks have been a difficult to deal with, especially because the of the spoofed IP addresses. Hence, I am into the research in the field of IP Traceback. I have few initial ideas, which is based upon the existing the schemes. I would like to have the skitter map, inorder to understand requirements of the scheme, deployment issues etc. I want the maps basically to have an idea about the number of ASes, hosts, edge routers, number of different paths etc. [November 2003]

  • International Computer Science Institute (ICSI) Center for Internet Research (ICIR)
    Mark Handley uses skitter data to validate ideas for alternative routing schemes to BGP. He investigates the extent that peerings fall into the obvious catagories (non-transit peering, customer/provider, etc). Route views data implies the vast majority of peerings are customer/provider and that most routes traversing Tier-1 providers don't transit, but there are a significant number of exceptions. But route views data is nowhere near complete, so skitter data might give a more complete view of what is actually happening. [January 2003]

  • Iowa State University
    Thomas E. Daniels is interested in developing new network traceback algorithms based on strategic placement of network monitors and novel approaches to searching their observations. He wishes to develop improved search algorithms tailored to actual network topologies. [October 2002]

  • KAIST, Korea
    Kwangwook Shin studies efficient frameworks for peer-to-peer lookup services, which is called Distributed Hash Tables, such as CAN, Chord, Pastry, and Tapestry. Our framework gives DHT more scalability and more efficiency in data lookup path latency. To show how much path latency we reduce, we need real internet topology data from skitter. [November 2003]

    Yusung Kim: Logistical Networking( suggests globally scalable storage service using IBP (Internet Backplane protocol) that is according to end-to-end principle like as IP. We'd like to add a multicast fuction into Logistical Networking. It will be overlay multicast or application layer multicast. We try to design that the overlay multicast should be topologically-aware and scalable. To test our design of overlay multicast, we want to use the real global Internet topology with large number of nodes. Logistical Networking also has already deployed 220 nodes, but we'd like to do an experiment using more bigger size, several thousands or more. [August 2003]

  • Kent State University
    Hassan Gobjuka is working on network monitoring and would like to get access to AS topologies so that I can run me experiments on besides using simulation models like Waman and Power Law. My final aim is to come up with efficient algorithms to monitor the network (including links) with minimum overhead generated. So I need real topologies to test my algorithms. [November 2003]

  • K.K. Wagh College of Engineering, Nasik, India
    Abhinav Kothari and 3 MS students want to geographically map skitter data to better inform their traffic monitoringa project. [August 2002]

  • Kongju National University, Korea
    Chang-Yong Lee studied RTTs using ICMP in both LAN and WAN. His research revealed that there was a long-range correlation in WAN, but not in LAN. This suggests that there might be a collective behavior in WAN traffic. What I am interested in this project is to determine the "transition point", transition from no correlation to long-range correlation. That is, he wants to see how many, on average, intermediate routers do packets have to pass in order to exhibit the collective behavior. To this end, I need e2e RTTs. Further, data containing intermediate RTTs will be very useful since it can be used to investigate the correlation progressively. It, however, was not easy to get enough data to carry out my project further. This is due mainly to some administration problem since not many servers in WAN allowed me to send ICMP packets.

    In this preliminary investigation, to determine a long-range correlation, I use standard methods of timeseries analysis, such as sample variance, Hurst exponent, and power spectrum. Both sample variance and Hurst exponent are measures to find a long-range correlation in time domain, whereas power spectrum can be used to determine a correlation in frequency domain. Both are, of course, related. I also plan to use other methods such as maximum likelihood estimation for both short and long-range correlations. Once long-range correlation has been confirmed, I plan to adapt the methods in critical phenomena in the statistical physics to determine the transition point and related exponents. [February 2004]

  • Konkuk University, Korea
    Chun-Sung Lee studying a method for .KR DNS name server's stability on behalf of the KRNIC (Korea Network Information Center). [July 2003]

  • Libera Universita Maria Santissima Assunta, Rome, Italy
    Luciano Capitanio uses skitter data for network visualization research. The goal of his project is to develop and test a new technique for traffic visualization. [April 2003]

  • LIP6 Universite Pierre et Marie Curie, Paris, France
    Kave Salamatian is the coordinator of the main French research project on Internet measurement (the Metropolis project). He plans to use Skitter data to study French connectivity to the Internet as well as some Internet graph oriented research. [November 2002]

  • MIT Cambridge, MA
    Russ Cox The Vivaldi project is developing distributed algorithms for assigning coordinates to Internet hosts so that latencies can be predicted by distances between these coordinates.The skitter data will provide us with a large data set to help in the evaluation of our techniques. [February 2004]

    Ji Li along with his advisor Karen Sollins, plans to use the skitter data to construct a practical Internet topology so as to do some performance evaluations in the Region Project. The underlying premise of the Region Project is that the concept of a region should be a new architecture capability in networking. A region is an entity that encapsulates and implements scoping, grouping, subdividing, and crossing boundaries of sets of entities. It is a powerful tool for managing the increasingly complex demands on the Internet. [June 2003]

  • NARA Institute of Science and Technology (NAIST), Japan
    Masafumi Oe will use CAIDA data for making BGP peering topology map and will run a simulation on it. [December 2002]

  • National Chiao-Tung University, Distributed System and Network Security Lab, Taiwan
    Chou Kang-Hsien uses skitter data for simulating DDoS prevention. He can simulate the situlation using IP-traceback for mitigating DDoS. [March 2003]

  • National Communications System (NCS)
    David Fuhrmann is determining ways to use skitter traceroute data and the associated topological views to provide indications of overall Internet Health as it pertains to the customer. [March 2003]

    Inette Furey and Sounil Yu are working to develop a real time, synoptic view of abnormalities in traffic flow on the major networks. These researchers are looking to see if the skitter data can be useful towards the development of this "synoptic view."

  • National Tsing Hua University, Taiwan
    Qian Yan's plan for using the Skitter data is to test a new IP traceback approach to defend against DDos attacks.We need traceroute maps of real Internet topologies to simulate DDos attacks and validate our design. [June 2004]

    Chien-Hsing Wang is observing the similarity of routes in Internet. Because the nature of routing is dynamic, to observe the similarity of routes may be helpful to observe the abnormality of routing. [March 2004]

  • New Jersey Institute of Technology
    Edmund Hibbert conducts research on Internet topology. I have just began to look over the "Sampling Biases in IP Topology Measurements" paper written by Boston University professors. I am interested in learning about identifying sampling bias in measurements such as traceroute. I would like to know if I could get a copy of the raw routing datasets used in the paper, which was referenced from this site. [June 2003]

    Rob Margrie studies the stability and diversity of international IP connectivity to the United Kingdom to determine the presence or absence of critical peeringpoints. [March 2004]

  • North Carolina State University
    Phil Lowden uses raw skitter/topology data to benchmark and guide studies into providing cost-based versus policy-based routing for Internet hosts. [September 2002]

  • Osaka Institute of Technology, Japan
    Yukiji Yamauchi and the graduate / undergraduate students at the department of Information systems are working for the project to visualize the network connectivity and the health report of the internet from within our institution. In order to minimize the number of the probe packets sent from our site, we need the accurate topology map of the internet handy. By using the result of our project, any institution will be able to create its own internet weather map with minimum probe packets. [June 2004]

  • Princeton University
    Stephen North and David Dobkin have been collecting data of the type available from skitter for some experimentation into the dynamic nature of the Internet. They are not sure that their data is giving them and accurate picture of the net and would like to run some tests on someone else's dtat to see if they get similar results. They expect to run these tests sporadically as a check against their own methods of data collection. [April 2004]

    David Dobkin and Dan Dobkin have been collecting data of the type available from skitter for some experimentation into the dynamic nature of the Internet. They are not sure that my data is giving me an accurate picture of the net and would like to run some tests on someone else's data to see if they get similar results. They expect to run these tests sporadically as a check against their own methods of data collection.

  • Queen Mary College, University of London
    Shi Zhou and Dr. Raul J. Mondragon, his supervisor, focus on the research of the Internet top ology. They want to analyse the skitter data (ITDK0204) and compare with other data sources in order to par ametrise key topological properties of the Internet. Their goal is to introduce a realistic generator for s imulation models of the Internet power law topology. Their research is funded by the U.K. Engineering and Physical Sciences Research Council. [April 2003]

    Gengxin Zhang is setting up a simulation model for internet to check the topology, connectivity and traffic load. The result of this simulation will be used in the Saveguard project. [March 2003]

  • Renselaar Polytechnic Institute, Ithaca, NY
    Krishna Kumar Ramachandra is working on formulating an analytical model for P2P networks and wish to model the core network as a component of the model. I need to determine the average degree of the router network to choose an appropriate mathematical model. This can be partially obtained by analysing the router map. [April 2004]

  • Saarland University, Saarbrucken, Germany
    Stefan Bender and 2 MS students use skitter data to estimate the distance between hosts in the Internet. Host distance metrics can inform application layer routing in a gnutella network. [August 2002]

  • Shanghai Jiaotong University, China
    Weixiong Rao performs peer-to-peer related research. Current P2P routing is unrelated to topology. Even rough topology data is likely to help yield better P2P performance. [August 2002]

  • Stanford University
    Ferdo Ivanek used skitter data to teach MS&E237, Progress in Worldwide Telecommunications. AS Internet graphs would be useful as illustrations of traffic. Interpreting Skitter graphs may be an informative term paper subject. [June 2003]

    Prasanna Ganesan is attempting to understand the impact of network structure and latency on overlay networks and applications working on top of such overlay networks. We hope to achieve more "acceptable" and realistic models of internet topology using Skitter data. [May 2003]

  • Sussex University
    David Ellis uses skitter data to validate new IP traceback systems. [June 2003]

  • Technical University of Catalonia, Barcelona, SP
    Dr. Jose M. Barcelo is supervising a PhD student who wants to study Internet topology generators. His first goal is to understand better the evolution of Internet studing interconectivity of AS's, size,path lengths at AS level and router level,etc. [October 2003]

  • Technological University, Sophia, Bulgaria
    Dimitar Vasilev is researching core networks. I want to track the flow between the core networks pow ering Internet. I want to collect data and generate core Internet networks weather maps. [November 2002]

  • Tel-Aviv University
    Karine Barzilai-Nahon performs research concerning social, economic and technological aspects of access to information in the Internet. [March 2003]

    Omer Ben Shalom's plan for using the Skitter data is to analyze the RTT and path from various Skitter collectors to a group of prefixes belonging to about 10-20 AS that show multiple AS 'Atoms' (groups of prefixes that have shared a common AS path from anywhere in the Internet) to try and find if those AS 'Atoms' share RTT and path characteristics lending more substance to the atom entity, this will be part of Omer's MSC thesis on BGP atoms.

    Helping Omer with his research is Anat Bremler-Bar, a Doctorate student in the Tel-Aviv University who is doing her doctorate on BGP related subjects with the same guide - Prof. Yehuda Afek. [January 2003]

  • Temple University
    Alan Izenman found reference to the skitter Internet data in a paper by Tang & Crovella. He would like to try multidimensional scaling on the skitter data (assuming it's in usable form) for a graduate course that he is teaching on applied multivariate analysis. The data appear to be very interesting. [June 2004]

  • Texas A&M University
    raghav Dube is conducting research in the area of detection of the sources of Distributed Denial of Service Attacks (DDoS). A new router packet marking algorithm is being developed and the skitter data will be used to test the effectiveness of the proposed technique. [March 2004]

  • Tokyo University, Tokyo, JP
    Ahmad Suffian is at Tokyo University and performs most of his research activities (focusing on Photonic Network) at Japan's National Institute of Informatics (NII, ) under the supervision of Prof. Dr. Asano Shoichiro (Director, Infrastructure Systems Research Division. He is also involved in Super-SINET ( ), which is the optical network backbone developed by NII. He used the CAIDA Skitter Data for his "Doctoral Exercise Subject", where he was assigned to make reports on Measurement Tools (such as Sniffer, Ethereal, MRTG, Netstat ect.) under the supervision of Dr. Hiroshi Esaki ( Dr. Esaki is involved with WIDE - Widely Integrated Distributed Environment ( [September 2003]

  • Tor Vergata University, Rome, Italy
    Filipo Radicchi performs statistical analisys of complex networks structure and is trying to find communities on the web with a fast search algorithm. [May 2003]

  • Trinity College Dublin
    Eoin Lawless is using skitter data to make test cases for a simulation of large networks. [June 2002]

  • UMWOW: Digital Media Interactive
    Eric Rodenbeck intends to use the skitter data to investigate the narrative potential of Internet data streams. His investigations will take the form of graphic and interactive storytelling, using the data as a starting point to build experiences that delight visually as well as intellectually. The process of investigation is intentionally open-ended, in order to allow for the possibility for a different kind of storytelling, and to open up the potential for new and interesting forms. For example, UMWOW is currently working on a project to represent live stock market data as a series of puppets sliding up and down poles, with the various aspects of the different markets affecting each puppets' size, height, 'mood,' etc. The CAIDA data will serve as the foundation for a similar kind of investigation taking global connectivity as a theme. (Site requires Flash plug-in.)

  • University of Aarhus, Department of Political Science
    Jakob Linaa Jensen will use the data for explorative purposes mainly and hopefully extract some data about the democratic use of the Internet. The project description is outlined below:

    The focus of his project is the democratic consequences of the Internet. To be more precise: does the Internet provide citizens with new possibilities for information-seeking, debating and dialogue with politicians and authorities? In these days existing democratic institutions are experiencing severe pressure from different sides, not at least the ongoing process of globalization. There are widespread hopes, however, that the new network technologies can contribute to a radical revitalisation of democracy.

    This project seeks to challenge, explore and eventually revise these hopes. The main hypothesis is, that the Internet contributes to enhanced democratic possibilities, but first and last the democratic renewal is in the hands of ourselves. Concretely, he will explore cases of democratic deliberation and debate at the Internet and will explore the Internet-mediated contact between citizens, politicians and authorities.

  • University of California, Berkeley
    Weidong Cui (with Prof. Randy Katz) is conducting research on distributed monitoring for network failuare detection. Raw skitter data will be used for constructing AS relationship and analyzing AS path properties. [March 2004]

    Feng Zhou uses skitter data to help understand the behavior of structured overlay networks such as Chord, Pastry and Tapestry. I'm especially interested in how they perform in exploiting the locality properties of the underlying network. [May 2003]

    Z. Morley Mao: Traceroute is widely used to detect routing problems, characterize end-to-end paths, and discover the Internet topology. Providing an accurate list of the Autonomous Systems (ASes) along the forwarding path would make traceroute even more valuable to researchers and network operators. However, conventional approaches to mapping traceroute hops to AS numbers are not accurate enough. We would like to develop more accurate techniques to map IP addresses to AS numbers by correcting the mappings based on origin AS numbers using a topologically diverse set of traceroute and BGP data. [April 2003]

    Ben Y. Zhao, along with advisors Anthony Joseph and John Kubiatowicz, are working on the Tapesty project. Tapestry is an efficient, decentralized and fault-tolerant overlay network that provides location services and routing to applications. To date, they've built prototypes of several interesting applications on top of Tapestry, including Bayeux (an application level multicast system), Mnemosyne (a stegnographic file system), and Ocean Store (a global-scale distributed storage system). In addition, they've added extensions to the Tapestry network for more efficient wide-area routing (Brocade) and support for rapid mobility (Mobile Tapestry). Their evaluation strategy for most of thse applications and for Tapestry itself is to utilize large-scale simulations on a packet level simulator. The issue is that they are limited to transit stub topologies generated by GT-ITM. Existing topology data out there all do not have real latency measurements associated with them. That's where Skitter will be very helpful, by providing both real latency and topology information.

    Thinh Nguyen is working on a framework for streaming video from multiple senders simultaneously to a single receiver in order to achieve higher throughput and fault tolerance. Video streaming over best-effort packet-switched networks is challenging due to a number of factors (e.g. high bit rates, delay, loss sensitivity). Many proposed schemes use scalable and error-resilient codecs in conjunction with network protocol to deal with the time-varying nature of the Internet. However, these schemes assume a single fixed route between the receiver and the sender throughout the session. If the network is congested along that route, video streaming suffers high loss rates and jitter. Instead, it is conceivable to have a subset of "best" senders simultaneously streaming video to a single receiver in order to provide the required throughput and minimize packet loss and delay. Having multiple senders also combats the unpredictability of Internet congestion. If the route between a particular sender and receiver experiences congestion during streaming, then the receiver can redistribute streaming rates among other senders, resulting in smooth video delivery. In our framework, we propose a receiver-driven TCP-friendly protocol to coordinate simultaneous transmissions of video from multiple senders. Our protocol employs two algorithms: rate allocation and packet partition. The rate allocation algorithm determines the sending rate for each sender to minimize the packet loss, while the packet partition algorithm aims to minimize the probability of packets arriving late. [July 2002]

  • University of California, Davis
    Matthew Caesar is working on an architecture for IP Telephony services, and would like to incorporate wide area traffic characteristics to make my results more realistic. He plans to use information including RTTs, loss rates, etc. to different hosts around the globe. Matt will use these numbers to construct an accurate simulation of a wide area ip telephony service.

  • University of California, Riverside
    Yihua He studies Internet routing asymmetry over the commercial lines in addition to academic networks from NLANR. [March 2004]

  • University of California, San Diego
    Barath Raghavan is developing a authenticated source routing system, and would like to do measurements of routers within various ASes of interest. He uses skitter data to find router IP addresses with these ASes. [October 2003]

    Yuchung Cheng is simulating p2p networks based on real networks measurements. [December 2002]

    Craig Donner, Mike Green, Daryl Sterling, and Tom Molina are grad student in the CSE department at UCSD, and are using skitter data for a section of a group final project for CSE 222 (Communication Networks). They are doing a research survey on Internet performance, with the subtopic being the role the topology of the Internet (connections between various ASs, suboptimal paths, etc) plays. They are hoping I'll be able to glean information such has how changing topology has altered routes (making them either faster or slower), and even basic things such as the number of links in an end-to-end connection (i.e. as this increases the latency of a connection increases, even with high bandwidth/low latency pipes there is still a minimum processing overhead per link).

  • University of California, Santa Barbara
    James Tandon is studying tele-robotic control over the Internet, it is essential to know what kind of induced delay to expect. Skitter data will be used to determine what kind of latency a tele-robotic control system will potentially need to compensate for. Current study focuses on the additional latency due to encryption protocols, and the impact that it has relative to normal RTT. Methodology: Compare compute time for encrypting a real-time, telerobotic control packet (essentially a UDP/IP packet of ~50-100bytes) with the round trip times seen over the internet. Determine what conditions that encryption becomes detrimental to real time control, i.e. for how many network hops is encrypt_time << RTT? How common is it to see RTTs < ~10ms? Do faster cryptographic algorithms need to be developed or can the current algorithms be used for the short distances? Are there ways to compensate for significant gains in delay due to encryption that don't reduce the security that encryption provides? [February 2004]

    Christopher Kruegel performs intrusion detection on routing advertisments and checks their validity. [March 2003]

  • University of Cambridge
    Andrew Moore uses skitter data for two projects: firstly to test work done on algorithms for wide area server discovery algorithms which allow queries to be serviced which return groups of servers, which have mutual constraints on their location. The second is related to the first a co-ordinate based location system, which will be used to calculate the co-ordinate for the hosts the system indexes. We need to create a new service as hosts will be un-cooperative so we have to have a distributed service that passively measures metrics to the host actively probes them. and the generic webpage for his work: Euan's work is still in a more amorphous stage but incorporates "providing network weather to distributed content networks. [October 2003]

  • University Carlos III of Madrid, Spain
    Juan Fco Rodriguez Hervela uses skitter data to test BGP behaviour using BRITE + SSFNet. He will try to generate useful real topologies and we compare the behaviour of BGP under these topologies with a new BGP modification we are developing related to the IPv6 protocol as well. [June 2003]

  • University of Chicago
    Adam Kalai wants to test algorithms for estimating the parameters of random graph models on specific graphs. He is interested in measuring the quality of particular models on different data sets. Initially, he will start with the preferential attachment models and hopefully extend it to copying models as well. [December 2003]

  • University of Illinois, Chicago
    Robert Grossman is using the skitter data to test a new protocol called Data Space Transfer Protocol (DSTP) which will enable easy data publishing on the web

  • University of Illinois-Urbana-Champaign
    Hyuk Lim is working on establishing an Internet coordinate system for estimating network delay. The skitter data will be used to validate the Internet coordinate system. A part of this work was presented in ACM Internet Measurement Conference 03. [June 2004]

    Apu Kapadia: Mist is a routing infrastructure for privacy-preserving communication. We would like to develop a distributed version of Mist (see URL for older version description). To test the performance of the enhanced Mist, we would like to model Mist Routers as AS's. For any realistic simulation, we would need a good approximation of the current Internet-AS topology. [June 2003]

    Braden is using skitter data in his semester project is to analyze network data to understand how the topology of the Internet changed (if at all) during September 11th. Perhaps he will be able to identify particular parts of the network that were stressed as major news houses were flooded with requests. This will most likely involve comparing network readings from Sept 4th and Sept 11th. He plans on using VTK (Visualization Toolkit) for the rendering and C++ for the data processing.

  • University of Kent
    Mok Yeen Nam is developing suitable network models that can be used to investigate optimisation techniques for large-scale overlay networks like the Internet. This project will examine the mapping between topological graphs and Euclidean space that preserves the relationships between the nodes as accurately as possible. Therefore, he hopes to use data from the CAIDA project to collects raw Internet data for mapping and research purposes to ensure that the models represent real-life configurations. [July 2003]

  • University of Kentucky
    Kenneth Calvert An important problem in building overlay networks is to form overlay connections among nodes that are 'close' to each other, where 'close' is defined in terms of network hop distance or network delay. Several techniques have been proposed to approximate the network proximity of arbitrary nodes. One such technique requires hosts on the Internet to measure delay and hop distances to a set of 'landmark' nodes. These network measurements are then then used to assign Euclidean coordinates to hosts, such that the Euclidean distance between two nodes is a predictor some form of "network distance" (e.g. delay) between the nodes. We are building a layer to support structured computations in overlay networks, and wish to incorporate such functionality into the layer. Our research goals are:
    1. Verify results presented in others' prior work that used skitter data.
    2. Propose and verify new techniques that improve on those results. [March 2004]

    Aditya Namjoshi is implementing a routing API for GT-ITM graphs. He looks into BGP policies and their effect on routing and other Internet parameters. He will examine both policies topology related data. Skitter data allows him to generate some realistic policies for a topology. [January 2003]

  • University Konstanz, Germany: Computer Science Institute
    Collaborators in the working group, "Databases, Data Mining and Visualization," Mike Sips, Daniel A. Keim, Christian Panse, and Markus Wawryniuk aim to develop a visual representation of the Internet, which allow a simple identification of interesting structures, so, for example, information highways. Their field of activity is the exploration and visualization of large datasets, for instance, on data sets such as

    - stock indices of various companies over a large period of time,
    - cutomer buying patterns, or
    - telephone connections within a given time frame, or
    - visualizing massive graphs

    In addition we have extracted, for example, clickstreams from E-Commerce databases and present this graphs in some visual form. However, such graphs have typically some hundred nodes and some thousand edges. However, we would gladly like to try out our ideas in much more massive graphs. With the help of these visual representations we would like to be able to estimate the effects of different dynamic events in the Internet. Therefore in the first step we would like to extract from the graph a hierarchy of the Internet. Other steps are the application of different transformations, particularly wavelet transformations, on the graph. The ultimate aim is the development of a visualization tool for the Internet.

  • University of Limerick
    Abia Moloisane: The Internet has had a profound impact on society. Not only has the Internet's emergence sparked a rapid transformation of the business world but it has also enabled vastly different business models to evolve and thrive. The Internet, however, has not reached its true potential of operating efficiency. As a result of network delays, outages and growing bottlenecks that represent a poor Quality of Service (QoS) status, the benefits of the Internet have been compromised. Among some initiatives to address the QoS issues is a proposal for establishing clear standard units of QoS and carrying out Internet Tomography measurement to capture performance data to enable QoS decisions. Capturing of performance data from an end-to-end perspective provides means for service providers/carriers to assess QoS and adopt a proactive mode of operation in addressing QoS issues/problems. However, this process raises security and privacy concerns and hence a non-invasive Internet Tomograph! My performance measurement technique is preferred for performance data capturing. An non-invasive Internet Tomography measurement system has the goal of delivering long-term and short-term (even continuous), comprehensive QoS assessment/ measurements of IP performance in the core, across the edge and even extending into customer sites without impacting negatively on normal user traffic or violating security and privacy concerns in the process. Systems may be implemented in a wide variety of organised and ?unorganised? ways. Typically a system is designed to monitor what the end-user QoS experience is at different locations and has to construct a comprehensive network statistical performance picture relevant to the interested parties. A system would typically be characterised by a constellation of automated probing stations generating and receiving test traffic. These are strategically located/positioned in relation to the Internet areas being assessed and would include both wired and wireless networks sections of these areas. With enough probing stations a general Internet performance tomographic picture is constructed, and the QoS measurement picture along certain geographic routes as a function of time, day, month, etc. may be constructed. [October 2003]

  • University of Michigan
    Junghee Han working with Prof. Jahanian at the University of Michigan plans to use the skitter data for her research. She is currently studing redundancy and path diversity of the Internet routing infrastructure. To make the study more representative, she expects to use the destination lists developed by skitter project and conduct experiments using the list. [August 2003]

  • University of Minnesota
    Steven Damer is going to implement a simulator to look at results for the Pi Path Identification mechanism for defending against DDOS attacks ( I am going to observe the effects of various degrees of router compliance with the protocol, and the effects of varying the path length stored. [December 2003]

    Kuai Xu investigates the presence of exchange points from Skitter data to characterize inter-AS connections over the Internet. The forward IP paths recorded in Skitter project could provide us a great source to find a large number of IP addresses within the address blocks of exchange points. A traceroute measurement containing such IP addresses indicates two or more ASes connecting to the same exchange points. Moreover, the AS topology within an exchange point can be discovered if the data source has a considerable coverage. [March 2003]

  • University of New South Wales
    Tim Moors wants to enhance the dependability of communication networks through systems implemented in end-ystems. In particular, part of this project involves creating a streamlined traceroute that reduces the time and number of probes involved when tracing routes. The intention is to compare Skitter data with that produced by our streamlined traceroute tool, and improve the reproducibility of our research by using Skitter data that other researchers can also access. [March 2004]

  • University of Oklahoma
    Johgnyun Kim wrote a simulation to validate epidemic theoretical model of computer worms, however this simulation needs some real Internet topology data. We will use the public skitter dataset to simulate the worm spread speed and efficiency of defense policies. In addition, we are interested in obtaining skitter datasets for our research conducting a simulation on multicast routing satisfying various constraints for "realistic" topology in the simulator. [September 2003]

  • University of Salerno, Italy
    Umberto Ferraro seeks a view [even partial] of the Internet graph in order to experiment with some new dynamic graph algorithms. He would like to use the skitter data to evaluate the behavior of these algorithms as the Internet graph evolves. Algorithms mainly concern the minimum spanning tree and the connectivity problems. [October 2003]

  • University of Saskatchewan, Saskatoon, Canada
    Niklas carlsson uses skitter data to model and better understand the Internet from a number of perspectives. The data will help obtain statistics that characterize such global networks, and hopefully can help improve current network models. The goal of this work is to (i) identify and characterize a number of metrics that influence the performance of data dissemination, and (ii) based on these characteristics, evaluate a number of dissemination algorithms. [June 2004]

  • University of Technology Sydney, Australia
    Netmap Analytics possesses unique large-scale visualization and data mapping software. It is mostly applied to analyse business problems of various sorts, e.g. fraud detection and in marketing areas. It also has a background in intelligence analysis. To date, the technology has not been used to research aspects of Internet traffic data flows, etc. However, John Galloway believes it has the potential to help enrich understanding in this interesting scientific domain. He plans to utilise the NetMap software to pursue research into understanding patterns of Internet relationships and changes in those over time.

    Samson Lee and 4 PhD students aim to develop an open, policy based architecture for the management of next generation telecommunications networks. expecting that a comprehensive Open Architecture-based Telecommunications Management network framework can manage the network based on centralized policies and roles. [September 2002]

  • University of Texas, Dallas
    Mehmet Baysan and his group are trying to imrove efficient techniques to reduce the feedback implosion on server on many-to-one communication services. They improved some methods to solve this problem. They tested this methodologies in some sample topologies. But now they want to simulate this methods on some real internet topology. They are working on many-to-one communication systems so we need to use a tree topology. In order to receive better feedback they want to use topologies as large as possible. [April 2004]

    Sevcan Bilir studies the path characteristics of the Internet. That is, given two randomly chosen network paths between two (source,dest) pairs on the Internet, he is interested in the relation of these two paths with each other. He plans to do this study using both AS-level paths and router-level paths. For this, he needs an AS-level and a router-level map of the Internet (or a part of it or an older version of the map from 1990s should also be okay). Ideally he would like to have a map in which! all alias resolution problems are handled, if possible. Alternatively, he will try to apply currently known mechanisms to handle alias resolution by ourselves. [December 2003]

  • University of Washington
    Neil Spring studies the geographic properties of Internet destinations, with an eye toward developing a geography-based overlay routing scheme. Skitter traces to /28's include much more detail at this global level than he could hope to recover on his own. He also hopes to experiment with composing skitter traces with properties that he can measure externally, now that skitter's map is much larger than it used to be. [December 2003]

    Ratul Mahajan is continuing the Rocketfuel project at the University of Washington. The current goals are to understand the routing policies used by ISPs. For this, the skitter data would be useful to understand the traffic flow patterns. [November 2002]

    Krishna Phani Gummadi, Steve Gribble, Scott Shenker, Ion Stoica, Sylvia Ratnasamy and RamaKrishna Gummadi will be studying the effectiveness of various proximity methods in DHT based overlay networks such as CAN, Chord, Pastry and Tapestry. Their evaluation includes simulations and analysis of various heuristics on synthetic topologies as well as more real world topologies. Assigning latencies to links in synthetic topologies like the GT-ITM is an unsolved problem. Here, Skitter data could prove to be very useful in providing realistic latencies. [November 2002]

    Ratul Mahajan and Neil Spring are starting out a study of Internet maps. They will be comparing data collected by various mapping efforts to see in what respects they differ from one another. One of the question we might try and answer during this study is how complete Internet map is when data from all the maps are combined. Interested in both AS-level and router-level maps, Skitter maps would be one of the data sources they plan to look at. Ratul and Neil will be working under David Wetherall on this project at the University of Washington. They are doing a study of BGP misconfigurations, in which they do active probing to verify their results. For this they need to find an active IP within a prefix. The skitter data helps them in doing this by providing a list of active IPs to begin with.

  • University of Wisconsin, Madison
    Margaret Richey and Hetal Patak are interested in the discovery of dark spaces of the Internet. They intend to use skitter data as a starting point to discover the locations of these dark spaces. This work is under the direction of Prof. Paul Barford. [April 2004]

    Jim Gast is studying proxy placement and sizing issues for content distribution. He will try to create simple algorithms for minimizing the total traffic caused by a traced workload given actual BGP routing tables. Current Internet routing may be sub-optimal in many places. Jim's measurements will be able to compare the existing network to the theoretical (mythical) optimal. Jim will be doing the work under Drs. Paul Barford and Larry Landweber at the University of Wisconsin / Madison.

  • Ural State University, Yekaterinburg, Russia
    Viktor S. Grichenko a post-graduate student under the direction of Michail Vladimirovich Volkov in the mathematical faculty, algebra and discrete mathematics chair, scientific supervisor has the primary practical goal of his research to test some N.Weaver and J.Nasario ideas on active worms keeping in mind spatial features of Internet topology. For now spatial effects (frequently mentioned in J.Nasario's papers) are the most significant objection on possibility of Warhol/Flash worms. He is planning to model monolithic (simple) and modular (complicated) worm spread over Internet.

    Key factors:
    - exploit effectivity (could be several exploits)
    - worm size
    - sysadmin politics
    - scan algorythms
    - software diversity

    In modular worm case some wave effects are expected but it's generally unclear how modularity will affect dynamics. The results obtained on real data could be precise estimate on expected worm spread speed and efficiency of defence policies. The theoretical part will focus on the the set of effects preliminary named "relative topology". The main reason for RT is inadequacy of ideal fully-interconnected Internet model for "travelling worm" problem. The real-world experimental data will be useful for this part too.

  • Vrije Universiteit Amsterdam, Faculty of Sciences
    At Vrije University they have a cluster-computer with a lot of computing power. They want to use this computer for GRID applications. However, in order to do this, they need to study the Internet topology and additionally they have to generate Internet weather maps. Sandjai Bhulai and others will use skitter data to aquire the topology, and to make better predictions of the Internet weather map. Furthermore, they aim to test the algorithms in practice.

  • York University, Toronto, Canada
    Vladimir Blagojevic uses skitter data to explore the option of using real Internet topology traces instead of synthetic topology for a P2P Internet media streaming simulator. [July 2003]

Related Objects

See to explore related objects to this document in the CAIDA Resource Catalog.
Last Modified