[nycbug-talk] Africa Network Traffic
o_sleep at belovedarctos.com
Sat Oct 6 14:44:47 EDT 2007
On Oct 5, 2007, at 5:13 PM, Alex Pilosov wrote:
> On Fri, 5 Oct 2007, Isaac Levy wrote:
>>>> Where to start? I'm drawing a blank on this...
>>> What is the internet?
>> Right :)
> To clarify, if you missed the point. You started with a very vague
> question. You need to define each word in the "how much of the
> CONTENT originates in Affica, West Africa, or even Ghana", and then
> realize how complex (or simple, depending on definition) your
> problem is.
I think Ike is just looking for correlations he can use to get an
understanding of network traffic in Africa, he probably has a good
understanding of how the internet works. Although the internet is
going to have various levels that obfuscate this, it's probably not
normally distributed. I think the biggest trick to this project will
be making things discrete. It's like when working on grid computing
and you are trying to figure out why certain jobs run slow or trying
to estimate how long a job will run. You have the data from the
hardware but it's trying to find which metric correlates with your
job times, and this won't be 100% either, but you need find an
imperfect match that you are comfortable with and then put your faith
in that metric and make decisions based on it. So for Ike or Ike's
friend, he is going to need to keep his question vague and general to
make sure he stays on track, he will get imperfect data and
correlations but he will need to find one's that he will be
comfortable with, i.e. they can be explained to correlate to a
certain percentage of error and the rare situations where those
metrics might be wrong, they are also explained. If they can get a
few metrics that agree to some level, or can be explained when they
don't agree then he will have good research.
So how does he get web cache data? Does akamai show their stats,
albeit in reverse because you want their customer (source) rather
their customer's customer (us)? Are they even public? Are there
companies that are and can this data be compared with what percent of
the market in Africa they actually have?
The actual research data is best had from the science and technology
library at 34/Madison or if you know someone at Baruch or CUNY, they
have access to research databases like Factiva or Lexis-Nexis and a
great library with business research on 24th/Lex. I am sure some of
the High Falutin Columbia guys can toot their own horn as well.
More information about the talk