caching-namserver

Bob Goodwin bobgoodwin at wildblue.net
Tue Aug 22 00:14:06 UTC 2006


Bob Goodwin wrote:
> *jdow wrote:*
>> *From: "Bob Goodwin" <bobgoodwin at wildblue.net>
>> *
>>> **Ed Greshko wrote:*
>>> *
>>>> **Bob Goodwin wrote:
>>>> *
>>>> *
>>>>> **I've done "yum install caching-nameserver" and with a few crude 
>>>>> tests
>>>>> it appears that it might be working.
>>>>>
>>>>> Is there a file of cached addresses that I can look at and see that
>>>>> they are actually being cached?
>>>>> *
>>>>> *
>>>> **
>>>> Not that I know of....
>>>>
>>>> However, you can do a "host -v somehost.com" were somehost is equal 
>>>> to a
>>>> host name you've not visited.  The first time you do it you may see
>>>> something like:
>>>>
>>>> Received 194 bytes from 127.0.0.1#53 in 1291 ms
>>>>
>>>> But if you do it again, and it is working, you should see something on
>>>> the order of:
>>>>
>>>> Received 194 bytes from 127.0.0.1#53 in 1 ms
>>>> *
>>>> * 
>>> **Ok, I tried:
>>> *
>>>
>>> *host -v speakeasy.net *
>>>
>>> *First time:
>>>
>>> Received 109 bytes from 127.0.0.1#53 in 745 ms
>>>
>>> Second time:
>>>
>>> Received 109 bytes from 127.0.0.1#53 in 105 ms
>>>
>>> That would seem to indicate that it works?
>>>
>>> I'm not sure it's of much value considering the round trip time to 
>>> the satellite but I guess every little bit helps.  I was timing the 
>>> "Looking up google.com" or whatever at the bottom of the Thunderbird 
>>> screen and timing that.  It looked faster after the initial trial ...
>>> * 
>> *
>> No - that would include cached images and all that nonsense. Use the
>> commandline and something like "host nlzero.com", one I bet you have
>> never touched before. Google you have touched before regularly and
>> is probably still cached.
>>
>> Round trip to the satellite? Then the second one did not come via
>> the satellite. But the delay is incredibly long for coming from
>> something local. (The first one does look like a satellite hop round
>> trip, potentially. One hop is a touch over 1/4 second absolute minimum.
>> Add to that some overhead for communications reliability and the
>> satellite's processing time, also.)
>>
>> {^_^}
>>
>> * 
> *Yeah, the satellite introduces a bit less than 500 ms of delay for 
> transit
> time in addition to the usual system delays.  When I look at ntpq I see
> typically 600-700 ms there.
>
> I thought the 100 ms I'm seeing for the local dns looked long but I 
> have no
> experience other than this.  As it is it's till less than the 700-800 
> otherwise. Yes, I understand that google is frequently used and should 
> be in the local cache,
> that was merely an example, thought I made that clear.
>
> Bob Goodwin
> *
>
Now another problem, I rebooted the computer and /etc/resolv.conf
is re-written without the "nameserver   127.0.0.1" line and whatever
cache was collected is gone, so the local stuff has to be collected again.

That and now you've got me worried about the time, I'm still seeing 
130-150 ms,
much shorter than using the Wildblue dns server but perhaps it should be 
better?

Before and after for "host -v google.com"

Received 252 bytes from 127.0.0.1#53 in 917 ms
Received 252 bytes from 127.0.0.1#53 in 151 ms

Bob Goodwin




More information about the fedora-list mailing list