All of your questions are good. The server where the tests were conducted
did not use icmp. The time in millisec was obtained from dig which
calculates the response time to a question on it's own.
Also the data is biased - the program is being run on an IP string which
is next door to ORSC ns1.diebold.net and you'll notice that system had a
the lowest minimum response time associated with it - that's because were
next door to it.
Also don't let the statement made by my junior confuse you - there's more
to the response time then meets the eye. In the case of non USG roots
most of them including the ORSC roots provide full answers, that accounts
for the greater time lag as the alternate servers do further recursion to
provide the full answer as opposed to the USG roots which only provide the
location of those roots associated with the question. Failure rates are
also governed by this difference.
Once the tests are completed - I intend to run the test program on a
number of different servers which are at a respectable distance from the
roots so a comprehensive average can be obtained. Also the test only
asked the roots 300 questions - which is respectable for a test - but the
full program in operation should ask at least 3000 questions of all roots
per day.
Let me know if you have further questions concerning this.
Regards
Joe Baptista
On Wed, 26 Apr 2000, @quasar Internet Solutions, Inc. wrote:
>
> No disrespect intended, Joe but seems we should have a little more
> background on how this data was obtained. For instance if you are pinging
> from your own server then the results are factoring in network conditions
> on your end as well. Then what we'd be looking at could be labelled
> "results of one average user somewhere at some given time".
>
> If you are pinging from a major intersection such as Mae East or Mae West
> then you will have a truer relative response time for overall purposes.
> That will differ as well depending on which. Time of day would be
> interesting to know as well. You may find you get significantly different
> response during regular daytime traffic, during prime time or during the
> dead of the night depending on how oversold someone's available bandwidth
> might be.
>
> Also it has been recently brought to my attention that some backbone
> providers (someone told me Time Warner does this) have something built
> into their routers that route ICMP traffic differently...i.e. that you
> should see what appears to be slower response times in results like this
> than is actual response for a "real" query. So I was a bit surprised to
> see that NS1.BEACHSHORE.NET appeared to be as close to NS1.QUASAR.NET as
> it did...it should 'appear' slower as it's primary provider is Time Warner
> where in actuality from what I can tell they are pretty close in
> actuality.
>
> Sorry if this reply has bored anyone as I would think that at least many
> on this list would already know this.
>
> -Dena
> -----------------------------------------------------------------------------
> -=Dena Whitebirch=- *** http://quasar.net/
> @quasar Internet Solutions, Inc. *** "Internet Powered by Experience"
> -----------------------------------------------------------------------------
> Supporting ORSC Root Zone *** It's only an address.
> -----------------------------------------------------------------------------
>
>
>
