next up previous contents
Next: Diffusion effect Up: Small world problem estimated Previous: Volumetric argument in more   Contents

Applying the formula

A homogeneous Gnutella is a fractal with dimension $ d=7.79$ according to the following table that lists some related problems estimated with our formula, in bold text we highlight the computed result, the other numbers are assumptions:


Description $ N$ $ M$ $ d$ $ l$
Gnutella $ 500000$ $ \mathbf{15.58}$ $ \mathbf{7.79}$ $ 7$ (TTL)
Milgram's experiment $ 250 \cdot 10^{6}$ $ \mathbf{24.38}$ $ \mathbf{12.19}$ $ 6$
World $ 6 \cdot 10^{9}$ $ 24.38$ $ 12.19$ $ \mathbf{7.78}$
Switzerland $ 7 \cdot 10^{6}$ $ 24.38$ $ 12.19$ $ \mathbf{4.47}$
Poschiavo $ 3500$ $ 24.38$ $ 12.19$ $ \mathbf{2.39}$
CPU $ 3 \cdot 10^{6}$ $ 3$ $ 1.5$ $ \mathbf{27257}$
Brain $ 100 \cdot 10^{9}$ $ 10$ $ 5$ $ \mathbf{218.67}$

For Gnutella, we know $ l$ as the standard TTL8 and the number of nodes computed by the Gnutella crawler[23]. Using the formula, we compute the dimension $ d$ and the average number of outgoing connections $ M$.

Milgram's experiment showed that in the United States, a country with about 250 million ($ N$) inhabitants, there are 6 ($ l$) degrees of separation. The formula estimates that each person knows about 25 people enough well to perform the experiment.

Using Milgram's $ M$, we compute $ l$ for the entire world, for Switzerland and for a little village in the mountains.

We try the formula on the brain, a complex network with 100 billion neurons. Each neuron has about 15000 connections but is connected to a neighborhood of about 10 other neurons only, about 1500 connections for each neuron. The path length would be then 218. We idealized the CPU as it would be composed by 3 million NAND ports, with 2 inputs and one output.

However, Gnutella and the above mentioned problems are far from homogeneous in the degree of the nodes, so that our formula gives only a rough estimation.


next up previous contents
Next: Diffusion effect Up: Small world problem estimated Previous: Volumetric argument in more   Contents
Tiziano Mengotti 2004-03-27