Setting up a Root name server

[chris]

Jim Reid wrote:

> >>>>> "chris" == chris  <chris at megabytecoffee.com> writes:
>
>     chris> We already have lots of local servers, we have 6 servers, 4
>     chris> authoritative and 2 resolvers. I would like to do
>     chris> everything I can to speed up the resolvers and provide our
>     chris> customers with the fastest possible service.
>
> This is not "a lot of local servers" for a network as big as yours is
> claimed to be. Especially if you've got 250,000 or so resolvers
> pointing at just those 6 servers. Maybe this is where your DNS
> problems really lie?
>
>

Woh, I said 250,000 ip's I'm sure some of them arn't use or are vhosts on
web servers, I'm not sure of the actual number of machines pointed to my
servers, and I doubt I will ever find out.



>     >>  This is just silly. Perhaps you'll configure your name servers
>     >> to slave every domain known to the DNS "to provide instant
>     >> response".  If you think you need to do this, you have other
>     >> deep-seated problems in your net. [As well as a poor
>     >> understanding of how the DNS works.]  Nobody else in the world
>     >> needs to slave .com, .net. etc (or feels the need to slave
>     >> them), so what makes things different for you? Really.
>
>     chris> Slave? Who said anything about slave? I see the word
>     chris> "Authoritative" in that paragraph.
>
> Slave is the new term for "secondary". This is what's used in BIND8 to
> define a secondary name server. BIND8's been out for over two years
> now and most people who run it use that definition. Nice to see you
> keep up to date with DNS developments.
>

Right, the root files are loaded via ftp and bind loads the zones as a
master, not a slave. If I was to "slave" off of a root server, then I
would be set up as a secondary and bind would have the lovely task of
named-xfering a gig or so of zone data.

>
>     >>  You've not identified the performance problem, far less
>     >> indicated how
>     >>
>
>     chris> There isn't a problem.
>
> So why are you saying there is? You're the one who claims that if you
> make your DNS lookups faster (for some unspecified definition of
> faster) your users/customers will see the difference. As others have
> tried to explain to you over and over, a few ms difference in DNS
> lookups will not be noticed. This is clearly a dialogue of the
> deaf. If you don't want to listen, why not shut up until you've done
> what you want to do and proven that it really makes the big difference
> that you mistakenly believe to be the case. I'm sure other big ISPs
> and companies would be interested in your findings.
>

Originally I posted a message here to find out some more information on
how to set up a root nameserver while in the process of setting it up. You
then decided to not answer my question, but to question my question, the
typical action of someone that doesn't know what they are talking about
but loves to argue. I thought this was a bind discussion group, not a
forum for bind related arguments.

I never said there was a problem, I came in to the project with the
intention of making a more optimal network. As I've said over and over
again.

Personally, I have heard of no one that has done this. I would like to try
it out. I don't have all the information I need to make it happen. So I
asked questions. If you don't like to read about people asking questions,
try unsubscribing from this list / news group.

>     >> remains: faster DNS lookups only matter when the existing DNS
>     >> setup is
>
>     chris> You are full of shit on that one. Faster DNS look up is one
>     chris> of the key things that makes an internet connection fast.
>
> We'll have to disagree on both points.
>
>     chris> I work with what I can control. I run the DNS servers here,
>     chris> and I would like to do what I can to make things as fast as
>     chris> possible with the resources available.
>
> So go and slave - secondary in your terminology - everything under the
> root domain. That'll make everything really fast for sure. In fact,
> why don't your run DNS for everybody on those servers so that
> everybody in the rest of the world can benefit from these super-fast
> name servers too.
>

5 years ago I would have been laughed at for running a caching nameserver
on my own lan.

>
>     chris> no, it's more like you type in nslookup, set server to a
>     chris> root server that you KNOW is geographically close to you,
>     chris> send it a query and get no response. Then you try another
>     chris> root server and still get no response. When this came up it
>     chris> was at about 6pm on a week day.. and I'm sure the internet
>     chris> was really cramped at that hour. But, the fact still
>     chris> remains, if I can get a root server on my network, and get
>     chris> our local nameserves to query it, and have our customers
>     chris> nameservers query it, it will speed things up. I don't know
>     chris> how much it will speed things up but I would like to find
>     chris> out. Since no one seems to have done this before.
>
> Because, as you've already been told several times, it's an utterly
> pointless exercise. A good name server will query the root name
> servers very infrequently so a "long" lookup will happen perhaps once
> or twice a day. So any performance improvement would only be for
> perhaps one lookup per TLD per day. Other lookups for that TLD will
> just get returned what was previously cached. They won't see any
> difference in response time irrespective of whether your name server
> was authoritative or not for the root or TLD zone.
>
> There's clearly no point in continuing this discussion. You don't want
> to listen or consider what's being said, so I for one have no
> intention of talking to you about this again.

Perhaps if you had spent more time talking about the problem and not
trying to cut it to shreds we could have saved everyone a lot of time and
bandwidth.