Re: caching-namserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



*Tim wrote:*
*On Wed, 2006-08-23 at 13:39 -0400, Bob Goodwin wrote:
*
*I have observed instances where the initial response time was near 1000 ms,
the second in the tens of ms and then subsequent checks would be above 100 ms?
*
*
Same here.

*
*Initially for host -a google.com:
Received 220 bytes from 127.0.0.1#53 in 1047 ms

Then a few seconds later:
Received 220 bytes from 127.0.0.1#53 in 38 ms

Now some hours later:
Received 156 bytes from 127.0.0.1#53 in 467 ms

Then using dig <dig google.com>
;; Query time: 168 msec
;; SERVER: 127.0.0.1#53(127.0.0.1)
;; WHEN: Wed Aug 23 13:15:23 2006
;; MSG SIZE  rcvd: 148

Repeated several times, always 164-168 ms.
*
*
I've noticed similar sort of things, myself.  Just to be clear, it
doesn't matter what made the first query (host, dig, etc.).  Once
something's looked up an address in your nameserver (e.g. web browser)
it can cache the results for the next thing (e.g. dig).
*
*Yes I understand that.
*
*

*
*Then here's one I have not visited today, an address Joanne provided the other day.

dig nlzero.com
;; Query time: 3180 msec

dig nlzero.com
;; Query time: 244 msec

Then the same address using <host -a nlzero.com>

Received 122 bytes from 127.0.0.1#53 in 382 ms

The correlation between the two methods is not perfect?
*
*
Various factors are going to mean there's variable delays between
results (other things your PC is doing at the time, it's not a real-time
system, and whether a thing works through several other things, in
sequence, to get its results).

Query goes out (takes time), DNS checks for local data (takes time), DNS
server may fetch remote data (takes time), DNS server responds (takes
time), querying program acts on data (takes time).  Quite probably the
more data returned with a record (i.e. several A records) the longer the
time reported by dig, though I would expect other things to be a more
significant variance to the time.  And certainly, if it has to go
through several remote DNS servers before finding the responses (queries
root server to find authoritative server for a domain, which then turns
out to not be authoritative and refers onto another server...), that'll
take longer.

*
*Yes, but I was getting opinions that the response time should be in the single digit range and that got me off looking for improvement. I was quite happy with the results I got initially after installing "caching-namserver." And I guess
I still am, it's doing what it's supposed to do ...*
*I did try pinging google with the Dell XP box yesterday.
The average of three pings was 775 ms
*
*
That's a different thing, altogether.  You're measuring the time it
takes for ICMP (usually) traffic to get between their system and yours,
DNS data may or may not be obtained from the same systems.  And DNS
traffic is usually UDP (a different scheme, probably with different
timing issues - depending on all the software systems involved from one
end to another).

e.g. Here I have several PCs, a few of them have webservers, a few of
them have DNS servers.  But it's most likely that if you pinged one of
them, some other PC's DNS server would have provided you with IP address
to ping.

*
*but I don't know how to check the dns response time with Windows?
*
*
There are dig type of tools for Windows, but I don't know a name to give
you a search parameter.  A quick search of "dig for windows" produces
some results, but I know nothing of them to give any recommendation
(other than to be cautious).

However, remember that, I think, you're mostly looking up the response
time of the queried DNS server, not the response time of the machine
asking the question.  If you're not running a DNS server on that Windows
box, you're probably not going to see the sort of results that you're
thinking about.
*
*It appeared that my dns response times were poor and I was beginning to think it was just this old computer. I knew it would ping 127.0.0.1 in under 100 µs so I tried the same thing on the new Dell XP box and found that Windows ping won't display anything smaller than 1.0ms. So there was no common ground there, that's all I was doing, it seemed there might be an observable difference between
a 450 mHz and a 2.8 gHz machine.  There probably is but I couldn't make any
meaningful measurements on the Windows computer. I really should put a Linux partition in it but it means buying an SATA drive for a computer that I don't
use often and it already has more space than I am likely to need.
*
*Many thanks for the help.

BobG

*


[Index of Archives]     [Current Fedora Users]     [Fedora Desktop]     [Fedora SELinux]     [Yosemite News]     [Yosemite Photos]     [KDE Users]     [Fedora Tools]     [Fedora Docs]

  Powered by Linux