[pim at bit.nl: Re: [dns-wg] DNS hardware]

Hi,

Sorry for duplicates, I group-replied and the reply seems to have
went to dns-wg-chair. Here's my original response to Bobby's question:
  
| I have a question that concerns the hardware in the server that should 
| be a primary DNS (even the hardware on the secondary DNS could be nice 
| to know about).
| What I wonder most about is, should the primary DNS be build with a 
| RAID-system? Or is that not importent of the fact there is a secondary 
| DNS?

It depends on the software you will be running and the amount of
queries per second you plan on answering.

In the case of bind(9), disk access is low because the server loads
zonedata into memory. Other software products (tinydns, nsd) use 
mmap()ed files, so they access disk but I don't think it'll be that 
much, because those pages are mapped into memory if you have enough of
it.

For the ISP I work for [medium sized ISP in the Netherlands], we have
two diskless P4 boxes and they each answer some 100qps. They are not 
loaded or busy at all, and happen to also serve as DNScache [using 
djbdns stuff at 500qps] and an SNTP service. 

nsauth1 up  139+08:53, load 0.17, 0.22, 0.23 FreeBSD/i386 (4.10-RELEASE-p5)
nsauth2 up  139+09:09, load 0.03, 0.02, 0.02 FreeBSD/i386 (4.10-RELEASE-p5)
nsauth3 up  246+14:55, load 0.05, 0.08, 0.07 FreeBSD/i386 (4.10-RELEASE-p5)

The difference in load is because most of our customers use nsauth1 as
caching nameserver and nsauth2/nsauth3 hardly ever gets hit.

groet,
Pim
-- 
Met vriendelijke groet,
BIT BV / Ing P.B. van Pelt
PBVP1-RIPE (PGPKEY-4DCA7E5E)