|
 PARCHER:
Real-Time Simulation of Global Internet Dynamics
It is widely
recognized that traffic in any Internet's link is highly fluctuating. The
analysis of fluctuations have revealed that self-similar nature of
Internet traffic. In particular, signal burstiness is extremely persistent
at very large resolutions. We have explained the origin of self-similarity
in Internet traffic by invoking the presence of phase transition
phenomenon (Solé and Valverde (2001), Valverde and Solé (2002)). The
Internet is modeled as a network of hosts or traffic sources that control
their own rate of packet injection and routers that store and
forward packets. In a recent paper (Valverde and Solé, 2004), we have
enhanced considerably our previous
models by introducing a realistic Internet topology generator (Yook et al,
2002).
Figure: An
snapshot taken from our realtime 3D simulator of global packet dynamics. Routers and Host are
displayed according to their (generated) geographical locations (see Yook
et al., 2002). The queue length at a node
grows and shrinks dinamically depending on the amount of incoming traffic.
Warm colors signal heavy packet loads. Notice how highly connected nodes (hubs)
are more likely to get congested.
| |
To download PARCHER application click here
(it works under
Windows 98, 2000 and XP).
This application requires OpenGL and a 3D graphics accelerator in order to
run smoothly.
|
|
The PARCHER
simulator fully implements the traffic model described in (Valverde and
Solé, 2004). Our model has the following key features:
Realistic Topology: Our previous analyses do not take into account the
heterogenous architecture of Internet. We have integrated the topology
generator described in Yook et al. 2002, which reproduces the global
Internet topology very well. They have show how most existing
Internet generators fall in a very different region of the phase space
where real Internet is located.
Adaptative Traffic Sources (hosts): Sources inject packets depending
on their perceived level of local congestion. This simple rule is enough
to self-organize the system in the critical point where a highly efficient
regime of traffic flow is achieved (Toroczkai and Bassler, 2004). At this point, the impredictability is also
maximized (see Solé and Valverde, 2001) and the system displays a great
diversity of behaviors.
Realistic router limitations: There is a limit in the maximum number
of packets a router can store. Packets arriving at full queues are
discarded as in the real Internet.
Path
horizon: Routing tables can not store detailed information about the
entire system. This will necessarily introduce an amount of disorder in
the paths taken by packets when travelling from one host to another. We have explored this cost/efficiency
trade-off in this model by introducing a parameter defining the visibiliy
scope of the router (or depth-of-routing paramter). This is closely
related to the size of routing table. When depth-of-routing equals the
network diameter the flow of paquets is maximized (see Valverde and Solé,
2004). Moreover, if the underlying network topology is small-world, then
the previous assignment is optimal in terms of the required amount of
routing information per node.
References
R. V.
Solé and S. Valverde,
'Information Transfer and Phase Transitions in a model of Internet Traffic',
Physica
A 289 (2001), 595-695
S.
Valverde and R. V. Solé,
'Self-Organized Critical Traffic in Parallel Computer Networks',
Physica A 312
(2002), 636-648
S.
Valverde and R. V. Solé,
'Internet's Critical Path Horizon',
European
Physics Journal B, 38(2), pp. 245, March (2004)
S.-H
Yook, H. Jeong and A.L- Barabási,
'Modeling the Internet's Large-Scale Topology',
Proc. Natl.
Acad. Sci. USA 99, 13382 (2002)
Z.
Toroczkai and K. E. Bassler,
'Jamming is Limited in Scale-Free Systems',
Nature, 428
716 (2004)
| 
