Etomic - Visualization

The network traffic visualizer tool

Start application - recommanded sun-java-jre 5 or higher

The ETOMIC Tomography Movie

István Csabai, Péter Hága, Péter Mátray, Gábor Simon, József Stéger and Gábor Vattay

Dept. of Physics of Complex Systems, Eötvös University, Budapest, Hungary

{csabai,haga,matray,simon,steger,vattay}@complex.elte.hu

Since the 1970s, the Internet has evolved rapidly from a few interconnected local area networks to a complex global network of routers both wired and wireless. This network has rapidly become embedded in the fabric of society. People use it for everything from email and web browsing to file transfer and Internet telephony. And its importance is likely to increase. Thus, in recent years, the study of the Internet and the way it behaves has taken on an increasing importance. One of our quests is to study the temporal properties of the Internet, such as the way patterns of congestion change in time.

We introduce measurement and evaluation techniques of network tomography to reveal the dynamic properties of the European Internet: the delays experienced by data packets as they move across the network. We study the distributions of queuing delays over a large part of the Internet in Europe and draw a map of network congestion. We are able to analyze the spatial structure of the Internet and identify highly congested or faulty segments.

We reconstructed the daily life (21. March 2006.) of the Internet and demonstrate the variation of the queuing delays in a short movie. Geographical positions of network routers and measurement hosts are well located on the map. Larger disks represent Etomic measurement nodes, whereas smaller disks stand for branching routers. In between them the resolvable network paths are drawn colored against the traffic experinced by the measurement probes. Green edges show lightly congested links and the larger the delays and their fluctuations the more red the colour of the edges. At first the overall congestion pattern is shown then later on a few interesting links are picked up and their unique queuing delay distributions are presented, too. We show the daily periodic behaviour of the averages of the queuing delays, which teach us there are two peaks during a day well connected to human habits: one noon time and one in the evening. (Play movie)