Network (Internet) as a Large System:
(Can we simplify the complicated large networks? How? In what sense?)
Do Young Eun
(Dept. of ECE, North Carolina State University)
Ness B. Shroff
(Dept. of ECE, Purdue University)
Publications
-
Do Young Eun and Ness B. Shroff, "Network Decomposition: Theory and Practice",
submitted to IEEE/ACM Transactions on Networking, Jan. 2003
(accepted for publication)
[full paper available upon request]
- Do Young Eun and Ness B. Shroff, "Simplification of Network Analysis in Large-Bandwidth
Systems", in Proceedings of IEEE INFOCOM, April 2003
-
Do Young Eun and Ness B. Shroff, "Network Decomposition in the Many-Sources Regime",
Advances in Applied Probability, Sept. 2004, To appear.
Research Statement and Questions:
One of the main characteristics of the current Internet
(or any other networks) is that it is "large" in many aspects.
At first sight, the largeness could mean a large capacity, as at the
core of networks, or a large number of flows multiplexed.
In this case, our network
decomposition techniques provides a framework in which
we can simplify the large network in terms of its performance metrics.
However, as in the current Internet, we can also find largeness in a
topological or geographical sense, such as the number of links attached
to a node, etc.
Internet itself is large and at the same time, highly asymmetric. It
is also robust in some sense, but vulnerable to organized attacks.
Its largeness inherently defies any exact analysis in most cases.
However, the largeness can also results in a simple analysis, giving
us rules of thumb instead of the complicated, optimal (in some sense)
solutions. Many approaches to this problem are available in the literature,
ranging from mathematical analyses based on probabilistic limit theories,
to an interpretation of large system using tools from modern physics.
- "Invariance principle" for large networks?
- Network dynamics and control
- Largeness in wireless networks?
- Does the large network topology affect network performance, e.g.,
QoS routing, or end-to-end QoS metrics? How?
Some Related Papers:
Do Young Eun's home page