Frank and Curtis: Can we get back to the bean counting? ("double every 100 days")
Frank said: The ESnet stats from the DOE [uplink from here] are impressive. Is the ESnet profile a part of the Public Internet that we're counting in our little bean contest?
Curtis replied: Frank-- ESnet is a closed internet but its basic characteristics are very similar to the large ISPs that we discuss. Because this is gummit supported, you can see the traffic growth statistics.
I went to the ESnet web site and got the 106 monthly octets from Jan 90 to Oct 98. I put them in an excel spreadsheet and calculated percentage increases month-to-month, every three months, every six months, and year-to-year. I did this as a rolling average, so for example there are 105 month-to-month comparisons, and 94 year-to-year comparisons, depending on the month you wish to start counting. I did this to see if I could find evidence of the doubling every 100 day claim in any 3,6,or 12-month period. Since 100 days is about 10% larger than 3 months, I was willing to accept growth of 90% every 3 months, or even 85% with a fudge factor. For 6 months, I chose an arbitrary threshold of 350%, and for 1 year I chose 750%.
I only found three 3-month periods with growth above 90%, and no 6- or 12-month periods. Furthermore, the 3-month high growth periods were at the beginning of the network when one would expect the largest growth: Jan-Mar 90 (163%), Feb-Apr 90 (95%), and May-Jul 90 (96%).
Now, it is possible to play games by taking single-month growth stats and extrapolating from them to 100-day figures, and if you do then you get 9 (out of 105) months in which the growth is at the rate of doubling every 100 days. The last such increase was Jun-Jul 98 (22%).
What happens, however, is that the growth is not sustained, and some months show negative growth.
Interestingly, the yearly stats fairly consistently show a doubling-every-year phenomenon, and the overall average of yearly growth rates is 99.2%, almost perfect doubling.
So what does this mean? I don't know other than ESnet does not demonstrate the claimed increase. OTOH, we should not expect ESnet to carry much e-commerce, so if it is only doubling every year, and it does not carry much e-commerce, then the e-commerce traffic has to be elsewhere, and perhaps a lot of it is going through UUNet, where the 100 day claim originated. Of course, if they extrapolated, or if they selectively chose a few quarterly growth stats, then almost any growth trend can be established. You can even use ESnet to show that internet traffic is decreasing at 7% every quarter (Nov 96 - Jan 97):-) |
