| Time / Zeit |
Room / Raum |
Type, Speaker, Topic /
Typ, Redner, Thema |
| 4th
quarter of 2009 |
|
13 Oct. 2009
14:00–15:00
|
Auditorium 2
|
Thesis talk:
Robert
Buchholz
HAIR – Prototype
Implementation and Evaluation
The Hierarchical Architecture for Internet
Routing (HAIR) is a Clean-Slate approach to
Internet routing and addressing that addresses both scalability
and mobility concerns in today's networks. This talk concludes a
diploma thesis whose goal it is to learn more about the routing
system by building and deploying a prototype setup. We will
discuss design decisions, evaluation results and lessons learned
of this work.
|
| 3rd
quarter of 2009 |
|
15 Sep. 2009
16:00–17:00
|
TEL 1118/19
|
PGT:
Benjamin
Frank
Developing Efficient Ranking Algorithms for the
Oracle Service
The Oracle service has been proposed as a novel
ISP-client collaboration paradigm towards efficient
formation of p2p overlays. An end-user submits
a list of candidate peers to the oracle server. The response
of the server is a ranked list based on simple performance
heuristics assuming full information about the underlying
network.
The task in this thesis are three-fold. First to provide a
flexible and extensible platform for simulations and analysis
of the proposed Oracle service within the SSFNet
simulation framework. Second, to enrich the ranking algorithms
taking into consideration semi-static metrics such as
AS distances, router-level hop distances, access
and backbone bandwidths, as well as dynamic metrics such
as delays, link utilization, available bandwidth and path
diversity, among others. Last but not least to study the
performance of ranking algorithms for both the end-user as well
as the ISP operating the Oracle service and thus an
analysis of the impact different parameters have on the overall
performance of a ranking function.
|
|
25 Aug. 2009
16:00–18:00
|
TEL 1118/19
|
PGT:
Juhoon
Kim
An accurate classification and analysis of
eDonkey traffic
The major purpose of this thesis is to
gauge the applicability of two different methods of network
traffic classification in light of the accuracy. Furthermore,
the method which was found out to be more accurate is used
to evaluate eDonkey traffic. In this thesis, I introduce the
analyzer implemented with two different traffic classification
methods which are based on the deep payload inspection. Moreover,
I evaluate the characteristics of eDonkey traffic with regard
to the popularity of the transmitted content, the frequency of
the transmission of signaling information and data payload. The
analyzer implemented for this research identifies TCP
and UDP packets of eDonkey traffic as well as kad
traffic. Kad is a DHT-based peer-to-peer file sharing
network and it is integrated in many eDonkey applications. Thus, I
expect that this analysis covers almost all the traffic which is
produced by eDonkey applications in the test evironment.
|
|
11 Aug. 2009
17:00–18:00
|
TEL 1118/19
|
PGT:
Gregor Schaffrath
Network Virtualization Architecture: Proposal and
Initial Prototype
The tussle between reliability and
functionality of the Internet is firmly biased on the side of
reliability. New enabling technologies fail to achieve traction
across the majority of ISPs. We believe that the
greatest challenge is not in finding solutions and improvements
to the Internet's many problems, but in how to actually deploy
those solutions and re-balance the tussle between reliability
and functionality. Network virtualization provides a promising
approach to enable the co-existence of innovation and reliability.
We describe a network virtualization architecture as a technology
for enabling Internet innovation. This architecture is motivated
from both business and technical perspectives and comprises four
main players. In order to gain insight about its viability, we
also evaluate some of its components based on experimental results
from a prototype implementation.
|
|
11 Aug. 2009
16:00–17:00
|
TEL 1118/19
|
PGT:
Mercè Lucas
Subirats
Statistical analysis of data from real world
mesh-networks
In recent years, there has been an increased interest in
cross-layer design for wireless networks. The magic pod of such
a design is to expose interactions and dynamics between the
layers within the protocol stack and use such information from
different layers to optimize the overall performance.
But what static and dynamic informations are we able to
collect from different layers within a real world mesh network?
What kind of informations are we able to measure directly with a
certain accuracy and what information are just inherent.
The focus of this work lies on the statistical analysis
of data, collected from real-world mesh networks. How to
characterize a wireless link in a multihop mesh-networks is
still an open research question.
|
|
14 July 2009
16:00–18:00
|
TEL 1118/19
|
PGT:
Nadi
Sarrar
Implementation and Evaluation of an Opportunistic
Mesh Routing Protocol
Today's wireless mesh networks usually use
a single-path routing protocol, derived from wired networks.
Opportunistic routing exploits the multicast nature of wireless
networks to gain performance. The goal of this thesis was to
implement and evaluate an opportunistic mesh routing protocol.
This talk mainly presents the evaluation results.
|
| 2nd
quarter of 2009 |
|
9 June 2009
16:00–17:00
|
TEL 1118/19
|
PGT:
Daniel
Levin
TCP Mobility for Multi-hop Wireless
Mesh-Networks
TCP suffers serious performance
penalties in multi-hop wireless mesh networks. Accordingly, a
replacement is needed to ensure reliable transport within the
mesh. Modifying the network stack on the end user systems, however
is not a viable option. TCP connections going into
and coming out of the mesh could however be bridged transparently
to the clients connecting at its edges. To the end systems, our
system would present a normal, contiguous TCP flow.
Within the mesh however, the end-points of the TCP
session must be managed, and the protocol state must migrate with
users as they roam. In this thesis, we propose the implementation
of this mobile TCP migration system.
|
|
26 May 2009
17:00–18:00
|
TEL 1118/19
|
PGT:
Ingmar
Poese
The Oracle server: implementation and performance
evaluation
In this thesis, we present the design, prototypical
implementation and analyses of the Oracle service, with
respect to the primary goals of efficiency, scalability and
flexibility. The contributions of the thesis are three-fold.
First, it sketches the disciplines underpinning the architecture
of the Oracle server. Second, it presents efficient data
structures and multi-threaded programming techniques used to
boost the performance of the Oracle server. Third, it provides
an exhaustive performance evaluation study of the implemented
Oracle server.
The experimental results presented in this thesis are
supported by deep system and performance analyses that provide
evidence in support of the efficiency and scalability of the
proposed architecture. To demonstrate the flexibility of the
proposed architecture a series of experiments that involve the
prototype server and real clients is also presented.
|
|
19 May 2009
17:00–18:00
|
Auditorium 2
|
During the project the Historic IP lookup
tool was developed, which can read BGP routing
updates, blacklist information, information about dynamic
IP blocks, etc. and which can answer
queries for IP addresses with historic information
like originator AS, route changes, or being on
blacklists.
This is not a full project talk, rather it is a short
introduction into the use and features of the Historic
IP lookup tool.
|
|
19 May 2009
16:00–17:00
|
Auditorium 2
|
PGT:
Mercé
Lucas Subirats
Statistical analysis of data from real world
mesh-networks
In recent years, there has been an increased interest in
cross-layer design for wireless networks. The magic pod of such
a design is to expose interactions and dynamics between the
layers within the protocol stack and use such information from
different layers to optimize the overall performance.
But what static and dynamic informations are we able to
collect from different layers within a real world mesh network?
What kind of informations are we able to measure directly with a
certain accuracy and what information are just inherent.
The focus of this work lies on the statistical analysis
of data, collected from real-world mesh networks. How to
characterize a wireless link in a multihop mesh-networks is
still an open research question.
|
|
12 May 2009
17:00–18:00
|
TEL 1118/19
|
PGT:
Harald Schiöberg
A Failsafe Architecture for Mesh Testbeds with Real
Users
(Dry-run for Mobihoc S3 Workshop)
We build a research testbed at our campus
with real users, namely all the members of the university, to
bridge the gap between synthetic mesh testbeds and productive
mesh deployments. This requires combining the flexibility of
research-only testbeds with the high reliability of production
networks. We show how we bring these two contradicting design
goals together, given the degree of flexibility we demand calls
for the ability to exchange the implementation of possibly all
layers of the network stack. Given the experimental nature of the
environment, we expect frequent software failures. However, we
argue our failsafe architecture can still limit the impact of such
failures to a satisfactory level of user experience.
|
|
12 May 2009
16:00–17:00
|
TEL 1118/19
|
PGT:
Juhoon
Kim
NNTP Traffic Analyzer using Bro
As the scale of the Internet grows every
day, attempts to analyze and understand past Internet traffic
for the estimation of future Internet traffic are increasing. In
this paper we first summarize NNTP (Network News
Transfer Protocol), which was designed for reading and exchanging
news group articles, then we describe an implementation of the
NNTP traffic analyzer which identifies, collects
and analyzes the NNTP traffic among the captured
Internet traffic. Finally, we perform an NNTP traffic
measurement on two Internet traffic samples which have been
recorded with TCP Dump. The analyzer used for
this research is based on Bro NIDS (Network Intrusion
Detection System) which is a Unix-based open source software.
|
| 1st
quarter of 2009 |
|
24 Mar. 2009
16:00–18:00
|
Futurum (TEL 1414)
|
PGT:
Markus
Konrad
Reducing Pollution in P2P networks using an
Oracle System
P2P systems make up for the most significant
portion of total traffic in the internet. Pollution in P2P
systems accounts for up to 50% of the traffic generated by such
systems. It has been found that such a high level of pollution
mainly exists because of poisoning, while unintended insertion of
polluted files into the network being practically none existant.
Our approach therefore trys to limit the number of downloads other
users perform from peers that actively poison the network. We do
this by using a P2P Oracle system, similar to the one proposed
by Aggarval et al. for proximity. Instead of
sorting nodes by proximity our oracle sorts by reputation, which
a peer builds over time. Our approach can either be deployed
directly to an existing P2P Oracle, where the weighting of
proximity to reputation is done at the Oracle system itself, or
seperatly as a new service, in this case the weighting is done by
the querying peer. In our work we evaluated different reputation
systems under different scenarios, like varying user behavior or
different pollution modells. In this speech we'd like to give you
an overview of the results we got.
|
|
17 Mar. 2009
16:00–18:00
|
Auditorium 2
|
PGT:
Hugo
Matilla
Fairness and Market Based Techniques in Wireless
Sensor Networks
The aim of this thesis is to bring solutions
based on economic tools and methods to the power consumption
problem in wireless sensor networks. Power consumption is
extremely critical as direct user intervention after initial
deployment is severely limited. The three basic functions of
wireless sensor networks are sensing, computing and communicating.
Problems in sensor networks arise in these three basic functions.
In each of these three classes the main questions to solve are who
is going to carry out the tasks and when. For example it is needed
to select which node is going to sense and to whom this data is
sent, also whether this data is going to be computed to give more
robustness to the message and how to do it in the best time.
Answers to these questions can be made in order to decrease the
power consumption (our case) and others like quality of service
(delays, data veracity etc.). Recent research has proposed several
solutions to increase the lifetime of wireless sensor networks.
The goal is to study how fairness is involve on the lifetime of
this kind of networks, and find out if it is possible whether the
usage of fairness will increase the lifetime on Wireless Sensor
Networks.
|
|
27 Jan. 2009
17:00–18:00
|
Auditorium 2
|
PGT:
Doris
Schiöberg
A Peer-to-peer Infrastructure for Social
Networks
Over the last years online social networks have become
more and more popular. For example several million are now
participating in Facebook, one of the most famous platforms.
People store a lot of data within such networks. This entails
giving the data to a company as all these networks are server
based. This data can then be sold to other companies. As a
consequence no one knows anymore where his data is and what
has been done with it. To give users back the control over
their data, we propose an approach for building an online
social network using a peer-to-peer infrastructure. Our
approach includes encryption of the data. In this manner,
the data can be distributed, is not under the control of a
company, and the user has full control about who can access
this data.
Soziale Netzwerke im Internet wurden über die
letzten Jahre zunehmend beliebter. Beispielsweise hat allein
Facebook, eine der bekanntesten Plattformen, mittlerweile
mehrere Millionen Mitglieder. Die Menschen legen enorme
Datenmengen in solchen Netzwerken ab, was gleichzeitig
bedeutet, diese Daten einer privaten Firma zu geben, da alle
diese Plattformen serverbasiert arbeiten. Eine Konsequenz
daraus ist, dass praktisch niemand mehr weiß, wo
all diese Daten tatsächlich landen. Wir stellen in
dieser Arbeit einen Ansatz vor, mit dem die Kontrolle
über die Daten wieder an den Benutzer zurückgegeben
werden soll; ein soziales Netzwerk, basierend auf einer
Peer-to-Peer-Infrastruktur, in der die Daten
verschlüsselt werden. Durch dieses System werden die
Daten im Netz verteilt, werden der Kontrolle einer einzelnen
Firma entzogen und der Benutzer erhält die volle
Kontrolle darüber, wer Zugriff auf seine Daten hat.
Talk will be held in english.
|
|
27 Jan. 2009
16:00–17:00
|
Auditorium 2
|
PGT:
Benjamin
Michéle
Using Onion Routing in Well-Established
P2P Networks to Provide Anonymity
Conventional approaches providing anonymous Internet
access to end users rely on chains of proxy servers called
Mix-nets, that are based on a small set of static mixes. Some
disadvantages of these approaches are, for example, weak
resistance against blocking by government-level firewalls,
poor scalability, and issues related to traffic analysis.
The most well-known and established representative of this
class of systems is Tor. On the other hand, there are a few
new systems that try to circumvent these issues by using
peer-to-peer approaches, however, most of them lack a user base
of sufficient size, which is vital for any anonymity system to
work reliably.
In this thesis, The Anonymous Peer-to-Peer Proxy
(Tap) is presented, which addresses those issues.
Tap builds on the design and know-how of Tor, but brings it
to the peer-to-peer world. Instead of using a small set of
dedicated servers to provide the anonymizing overlay network,
Tap uses all participating clients as servers as well. This
approach is much more resistant to blocking due to the large
amount of daily changing nodes. The geographical diversity
of node addresses makes it more difficult for adversaries
to observe all participating nodes and therefore impairs
the ability to perform attacks based on traffic analysis.
Furthermore, it scales very well because each additional client
also adds his own resources to the network.
To address the problem of a small user base in new systems,
Tap is bundled with a popular peer-to-peer file-sharing client.
As users tend to adopt new releases rapidly, a network with
thousands of users could be established within a few weeks.
It also provides an excellent opportunity to replace Tor's
central directory servers, which are used to locate Tor routers
and their corresponding keys. Tap uses the file-sharing
client's mechanisms to get to know other nodes and the client's
distributed hash table infrastructure to look up their
respective keys.
|
|
20 Jan. 2009
16:00–17:00
|
Auditorium 2
|
PGT:
Carmen
Mañez
Analysis of Social Networks based on Traces of Internet
Chat Traffic (diplom thesis debut talk)
This thesis is about extracting the
communcation graph of ICR users from network traces. As a next
step a set of graph properties will be calculated from the graph
to be able to compare this social network with both (i) other
social networks in the online world (i.e., from
OSNs) and (ii) offline social networks (real humans). For results
from OSN friendship graphs we rely on a public MySpace data set
and results of other researchers. For the offline networks we
migth get access to a data set that is currently constructed here
at T-Labs by researches from LMU and can resort to result from
the field of social sciences. As a side result of the project
differnent graph visualisation tools will be explored.
|
|
13 Jan. 2009
16:00–18:00
|
Auditorium 2
|
PGT:
Sebastian
Mellmann
Evaluation of VoIP Quality of
Experience (QoE) under different networking
conditions
With the emergence of new services and fierce competition
among the Internet service providers, focus of researchers is
shifted from Quality of Service (QoS) to Quality
of Experience (QoE), i.e., quality as
perceived by the user. The main goal of this thesis is to create
Internet like dynamic conditions in RouterLab testbed and
evaluate QoE under various different networking
conditions for Voice over IP application. For this
purpose a software mechanism ExpAuto will be developed
which will ensure automation of experiments in the testbed
from a central point along with collection of experiment
data. Results of the experiments will be analyzed in order to
establish relationships between networking parameters such
packet-loss, delay, jitter and user perception.
Talk will be held in german! (Slides will be in
englilsh.)
|
4th quarter of 2009
2008's talks