On Wed, 29 Jan 2003 13:38:46 +0000, Daniel Lorch <[EMAIL PROTECTED]> said:
> Why didn't you just increase the TTL? The current TTL of 1800 is far
> too low, causing unecessary traffic and latency, as it does not stay
> long enough in the ISP's caches. It is common to use TTL's around
> 172800 (2 days), at least for static addresses. If your server is
> down for more than 2 days, you're out of business anyway.

Long TTLs are fine, but they also define how fast changes in the zone
(changes for existing names in the zone to be exact) propagate through
the Internet.

In some cases you want to be able to change a given name faster than
that; for example if a given IP address is under persistent attack
(such as the address of www1.whitehouse.gov a while ago) or is
otherwise DOSed.

There is a nice paper studying the impact of TTL values on DNS traffic
and response times: "DNS Performance and the Effectiveness of
Caching", by Jaeyeon Jung, Emil Sit, Hari Balakrishnan, Robert Morris
(yes, *that* Morris :-), Internet Measurement Workshop 2001 (IMW 2001):

http://nms.lcs.mit.edu/papers/dns-imw2001.ps.gz

The gist is that caching at the delegation points (NS records and
their associated A records) is what provides most benefit, while the
impact of low-TTL A records isn't that bad.

So if you want to be able to quickly modify things *within* your zone,
but you still want to benefit from DNS caching, use a smaller TTL for
most records (maybe using the $TTL directive), but specify larger TTLs
for the NS records and the A records for the names that the NS records
point to.

The swinog.ch zone is now configured that way.

Regards,
-- 
Simon.
----------------------------------------------
[EMAIL PROTECTED] Maillist-Archive:
http://www.mail-archive.com/swinog%40swinog.ch/

Reply via email to