I'm trying to design an inexpensive large scale DNS server but fail to find any metrics or methods to base scalabilty.

Can anyone offer information on building a stable dedicated DNS server? That might be able to scale well.

Thanks in advance,
-rg

You didn't indicate how large an implementation you're considering. However, Bind should certainly meet your needs, if not perhaps "overkill."

You may want to consider dnsmasq. It claims to be capable of handling DNS and DHCP for at least a thousand clients and, if the size of the DNS cache is increased: the hard limit is 10000 names.

My original post did mention the scalability I desired but it seemed to exceed the conversation threshold in LQ so it was removed/broadened.
(original post 1M+ entries per month, which is still the desired goal to meet/exceed)

Is that 1 million new name records that this DNS system will be authoritative for added each month? I can't imagine adding that many new records indefinitely, there must be some upper limit to the number of name records you will be resolving.

Or do you mean 1 million queries a month (which seems rather low in the grand scheme of things).

Either way I would think Bind should serve just fine, it will just be a matter of choosing the right number of servers and distributing them around properly to handle load and give redundancy.

That's correct, 1M new authoritative entries per month. I can't say as to when it will reach a limit but I am trying to plan for up to 100m entries.

Choosing the right number of servers is a factor of how much one server can handle... and we're back to the original question.

How many entries can one bind server handle? What are the server specs for a fully blown BIND implementation?

-rg

How are you coming to those numbers, and what will your business be that you think you will get that many domains using your DNS servers to be authoritative? I am pretty sure that by current measure there are currently just over 100M registered domains, so by you aiming to get around 100M registered domains that would mean either a huge spike in registrations, or taking over business from the large established registrars.

That aside I would venture to guess you could actually do this with a fairly small amount of modern hardware. A good starting point would be two authoritative servers which won't resolve to clients, and two or more resolving servers. You will also need to decide if you need/want to offer a general DNS service where clients could query the cache of your servers or you will only be providing resolution for domains you are authoritative for.

OK, maybe I misspoke. Let me be clear. Instead of 1M new domains per month, it's actually 1M new entries per month within a single existing authoritative domain.

That is, the server(s) must be authoritative for all entries (but they mainly exist within a single TLD.)

-rg

It still seems like a very large number of domains to be in charge of, especially if this is a new business. If this is new I would imagine there would be a ramp up period, in which you could start out with a smaller DNS infrastructure and then monitor the load and scale that up by adding servers as needed. I think the basic setup of BIND with two authoritative servers, and two resolving servers to start, will be a good foundation.

The current setup already has 80M entries (grew 20M+ last month) and the original question remains unanswered.

Note: 80M entries does not imply 80M domains; in this case, all entries are within one single domain.

Thanks for your input anyways,
-rg