RE: fqdncache_size help

Hi,

]From: Simon Dray [mailto:Simon.Dray@bbsrc.ac.uk]
]
]Dear all
] Stupid question probably but here goes, does increasing the
]size of the fqdncache_size allow you to store a larger amount of fqdn
]lookups therefore reducing the amount of time that a dns lookup takes ? if
]so are there any recommend sizes, I take that the lifetime of an entity in
]the fqdn cache is controlled by its ttl and that this is set at lookup time.

Our old cache hierarchy is split into the virtual domains !.com and .com.
The !.com domain service lookup times are critical, e.g. there are several
authoritative dns servers in more remote corners of the world with an
abyssmal reachability.

<personal opinion>
I had to increase the ip cache size in order to achieve reasonable ip
cache hit rates, 4096 for *.com and 8192 for !.com (still too small). But
my old caches are rather busy. Look into your mgr:ipcache header, and

        (100 * (Requests-Misses)) / Requests

should be well above 90%. If your

        (100 * Hits) / Requests

is above 90%, too, you should be fine. At least to the best of my
experience.

You might also want to draw the average/median service time and request
rate in some kind of Mrtg chart to watch out for glitches. The sequence of
maybe faulty logic is: you probably see peaks in the service time when
there is a high dns request rate. Both together mean that your ip cache
was not large enough to buffer away some peaks. So, if your average or
median DNS service time continously stays below some threshold dependent
on your ISP and internet connectivity, you don't need to worry about your
ip cache size. You'll be fast enough fetching ip cache misses.
</personal opinion>

Happy Caching,
Dipl.-Ing. Jens-S. Vöckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726
Received on Fri Feb 11 2000 - 09:03:42 MST