Past Seminars

 

The seminars listed below were organized by Dr. M. Karaliopoulos and Prof. I. Stavrakakis and took place periodically in the Department of Informatics and Telecommunications, University of Athens. Their scope extended over various research topics in the area of computer and telecommunication networks.

 

19.06.2012


 

Title: Performance Analysis of Opportunistic Networks under Realistic Mobility Patterns

Speaker: Thrasyvoulos Spyropoulos (Eurecom, FR)

Webpage : http://www.eurecom.fr/~spyropou/

Time : 11.00

Room:  Aιθουσα E

 

Abstract

Epidemic algorithms have found their way into many areas of computer science, such as databases and distributed systems. Recently, epidemic spreading and variants (e.g. gossip) have been proposed for communicating in Opportunistic and Delay Tolerant Networks (DTNs).  To ensure analytical tractability, existing analyses of epidemic spreading predominantly consider homogeneous contact rates between nodes. Nevertheless, numerous studies of real mobility traces reveal a different picture: contact rates between different pairs of nodes can vary widely, and many pairs of nodes actually almost never meet. This raises the question: can we derive useful and accurate closed form expressions for the performance of epidemic schemes, under more generic mobility assumptions?

In this talk we will start from the basic epidemic scheme and simple heterogeneous mobility models and progress our way towards more generic mobility scenarios and optimization algorithms.
We will show that for some classes of models, highly accurate closed form approximations can be derived based on simple markov chains, despite heterogeneous inter-meeting rates. We will then consider arbitrary contact graphs, and show that, while traditional MC-based models break, (spectral) graph theory provides us with the means to get useful approximations to complex processes over generic opportunistic networks.

Bio

Thrasyvoulos Spyropoulos was born in Athens, Greece, in July 1976. He has a diploma in Electrical and Computer Engineering from the National Technical University of Athens, Greece, in February 2000, and a Ph.D. degree in Electrical Engineering from the University of Southern California (USC), Los Angeles. In the past, he has been with the Planete project-team at INRIA, Sophia-Antipolis, and with the Communication and Systems Group at ETH, Zurich. He is currently an Assistant Professor at EURECOM, Sophia-Antipolis, France. He is the co-recipient of the IEEE SECON 2008 best paper award and runner-up for the ACM Mobihoc 2011 best paper award.


07.06.2012


 

Title: Incentive Mechanisms for Hierarchical Spectrum Markets

Speaker: George Iosifidis (University of Thessaly, GR)

Webpage : http://georgeiosifidis.net/

Time : 16.00

Room:  A56

 

Abstract

Nowadays, it is common belief that the current coarse and static spectrum management policy creates a spectrum shortage. While this resource is expensive and scarce, significant amount of the reserved spectrum remains idle and unexploited by legitimate owners A prominent proposed solution is the reform of the spectrum allocation policy and the deployment of dynamic spectrum (DS) markets. This is expected to increase spectrum utilization and already several related business models exist in the market. However, the market-based solution is not a panacea and should be carefully applied.

In this talk,  we will discuss spectrum allocation mechanisms in hierarchical multi-layer DS markets which are expected to proliferate in the near future based on the current spectrum policy reform proposals. We consider a setting where a state agency sells spectrum channels to Primary Operators (POs) who subsequently resell them to Secondary Operators (SOs) through auctions. We show that these hierarchical markets do not result in a socially efficient spectrum allocation which is aimed by the agency, due to lack of coordination among the entities in different layers and the inherently selfish revenue-maximizing strategy of POs. In order to reconcile these opposing objectives, we propose an incentive mechanism which aligns the strategy and the actions of the POs with the objective of the agency, and thus leads to system performance improvement in terms of social welfare. This pricing-based scheme constitutes a method for hierarchical market regulation and allows the network designer, the agency, to improve the efficiency of the hierarchical channel allocation.

Bio

George Iosifidis holds an Engineering Degree in Telecommunications (Greek Air Force Academy, 2000) and a M. Sc. in communication networks (University of Thessaly, 2007). He recently completed his PhD studies in University of Thessaly under the supervision of Prof. Iordanis Koutsopoulos. His research interests are in the area of network economics and network optimization with emphasis on overlay networks and spectrum management.

 

 

24.05.2012


 

Title: Scaling Laws for Joint Content Replication and Delivery in Wireless Networks

Speaker: Savvas Gitzenis (Center for Research and Technology (CERTH)-ITI, GR)

Time : 16.00

Room:  A56

 

Abstract

The scalability of wireless networks has gained significant interest since the seminal work of Gupta&Kumar that derived the pessimistic result that in large wireless networks with multihop communications where each user uniformly selects another node to communicate with, per-user throughput falls asymptotically as to the inverse of the square root of N, the number of nodes in the network. The intuitive explanation behind this law is that to reach the destination, messages should be relayed over a number of hops that is on average proportional to the square root of N, i.e., the network diameter.

In this talk, we switch perspective to the novel paradigm of Content-based Networking and study the new resultant scaling laws. In particular, in Content-based Networking, requests are placed for specific content, as opposed to a specific node, and, moreover, nodes are equipped with buffers/caches where the content is replicated; thus, the requested content may exist in multiple places in the network. Intuitively then, we aspire in breaching the previous O(N^.5) network diameter hop-count, and switch to more favorable scaling laws for the user throughput.

To this end, we start by formulating the detailed joint content replication and delivery problem, which is a hard combinatorial optimization. This is reduced to a simple replication density problem, that can be solved directly using KTT conditions/Lagrange multipliers. We show that the latter problem's performance is of the same order to the original problem, which enables us studying its scaling behavior. Assuming the Zipf popularity distribution regarding the content requests, well-noted in Internet traffic, we derive the associated scaling laws. In particular, the asymptotics are considered with respect to the number of nodes N of the network, volume of the content or number of files M, and the node cache capacity K. Letting them scale to infinity in various regimes, we derive laws ranging from O(N^.5), as in Gupta&Kumar, down to O(1), i.e., the best possible result.

Parts of this work have been presented in the INFOCOM 2012 and SPASWiN/WiOpt 2012 conferences.

Bio

Savvas Gitzenis has received in MSc and PhD degrees from the department of Electrical Engineering of Stanford University and his diploma/BS degree from the National Technical University of Athens. He is currently a researcher at the Institute of Informatics and Telematics of CERTH and has been involved in several FP7 research projects. Savvas has served as an adjunct lecturer in the Department of Computer and Communication Engineering at the University of Thessaly and previously has been a post doctoral researcher at Sun Labs in Menlo Park, California. His research interests lie in the area of resource allocation and distributed control in networks and computing systems, and in particular, in wireless networks and mobile computing, investigating on autonomous and scalable policies.

 


10.05.2012



 

Title : Internetworking with DTN

Speaker: Vassilis Tsaoussidis (Democritus University of Thrace, GR)

Webpage: http://www.intersys-lab.org/pages/members/vassilis-tsaoussidis.php

Time : 16.00

Room:  A56

 

Abstract

I will discuss the properties of internetworked systems that incorporate DTN (Disruptive-/Delay- Tolerant Network) technologies and claim that DTN has the potential to gradually become the network protocol of choice for all heteregoneous internetworks. I will highlight examples from the space and opportunistic network domains and present my view of Future Internet along with a corresponding research agenda.  During my talk, I will present the projects, activities and research achievements of the Space Internetworking Center.

Note: FP7 and ESA support is acknowledged.

Bio
Vassilis Tsaoussidis (Professor, ECE, DUTH) holds degrees in Applied Mathematics (Aristotle University, Greece) and Computer Science (Ph.D in Computer Networks Humboldt University, Berlin, Germany). Vassilis held a postdoctoral appointment at the Department of Computer Science, Rutgers University, New Brunswick, NJ and faculty appointments at the Computer Science Department of SUNY Stony Brook, NY and the college of Computer Science of Northeastern University, Boston, MA. He was Visiting Professor at MIT, where he served his sabbatical in 2009 and is a research associate since then. Vassilis is the Director of "Internetworked Systems Lab", and founder of "Space Internetworking Center - SPICE".

For more details, please check http://www.spice-center.org/

 

05.04.2012


 

Title : Quantifying location privacy

Speaker: George Theodorakopoulos (University of Derby, UK)

Webpage: http://www.derby.ac.uk/computing/research/dr-theodorakopoulos

Time : 16.00

Room:  A56

Abstract

The popularity of personal communication devices leads to serious concerns about privacy in general, and location privacy in particular. As a response to these issues, a number of Location-Privacy Protection Mechanisms (LPPMs) have been proposed during the last decade. However, their assessment and comparison remains problematic because of the absence of a systematic method to quantify them. In particular, the assumptions about the attacker’s model tend to be incomplete, with the risk of a possibly wrong estimation of the users’ location privacy.

I will talk about how we address these issues by providing a formal framework for the analysis of LPPMs; it captures, in particular, the prior information that might be available to the attacker, and various attacks that he can perform. By formalizing the adversary’s performance, we propose and justify the right metric to quantify location privacy. We find that popular privacy metrics, such as k-anonymity and entropy, do not correlate well with the success of the adversary in inferring users’ locations.

Joint work with R. Shokri, J.-Y. Le Boudec, and J.-P. Hubaux, appeared at the IEEE Symposium on Security and Privacy, 2011.

 

22.03.2012


 

Title : Fair Background Data Transfers of Minimal Delay Impact

Speaker: Antonis Dimakis (Athens University of Economics and Business, GR)

Webpage: http://nes.aueb.gr/dimakis.html

Time: 16.00

Room:  ΝOC Teleconference room, ΝΟC building (Department of Informatics and Telecommunications), 2nd floor

(alternatively : amphitheater A1, depending on attendance)

Abstract

In this talk I will present a methodology for the design of congestion control protocols for background data transfers that have a minimal delay impact on short TCP transfers and compete for a target share of the leftover average capacity with other background TCP transfers. The optimal policy is derived but as it turns out it is difficult to implement in practice. Fortunately, it can be well approximated by a weighted TCP policy, which maintains a target proportion of TCPs bandwidth at all times in order to achieve the same share of leftover average capacity. The relative approximation error is always less than 17.2% while it quickly vanishes as the number of coexisting background TCP flows increases.

Next, I will describe a general utility-based fairness criterion for sharing the leftover average capacity, including a penalty term capturing the delay impact to short flows. A distributed weight adjustment policy is considered where, at equilibrium, the overall performance is nearly optimal, with a vanishing relative optimality error as the number of background TCP flows increases. I will illustrate the methodology by giving two examples of congestion control algorithms for background transfers. Both achieve low delay for short flows relative to TCP, but at the same time they present strong incentives for adoption against incumbent low priority solutions in public environments.

This is joint work with Costas Courcoubetis and Michalis Kanakakis (to be presented in IEEE Infocom 2012