Persistent cache

Danny Mayer mayer at gis.net
Mon Nov 11 18:41:13 UTC 2002


At 09:56 AM 11/8/02, ObiWan wrote:

>Danny Mayer <mayer at gis.net> wrote in message 
>news:<aq61l0$c0m0$1 at isrv4.isc.org>...
>
> > For example, cache poisoning becomes much harder to get rid of and
> > the costs of the first lookup are going to be lost in the noise of
> > everything else the server is doing. It's almost not measurable outside
> > of a test.
>
>Well, first of all we hadn't a single case of cache poisoning
>till now, if you can give me some pointers to test I'll be happy
>to perform any testing on the above, regarding the timings I invite
>you to install/run a copy of BIND-PE and to test it by yourself,
>I'm not talking of "fried air", we made *tests* (and a whole bunch
>of them) and all tests demonstrated how in a desktop machine
>environment (a machine which gets shutdown often) persisting the
>cache data has a positive impact on performance (and I'm not talking
>about a two milliseconds gain here)

I'm not about to reinstall a copy of the old code. I've already run tests
against the 9.2.x code base, but not for what you are doing.

The problem with the tests that you are likely to run is that they are
just that: tests. They don't mimic the real world, particularly on a
desktop machine which is barely likely to issue one request a second
and that's when it's just getting started. I'd like to see numbers and how
they are improved by your strategy when starting up an application
and getting it ready for use.

Danny



More information about the bind-users mailing list