mem sizing for local vs. cached

Kevin Darcy kcd at daimlerchrysler.com
Tue Nov 21 23:28:00 UTC 2000


Mark.Andrews at nominum.com wrote:

> >
> >
> > > Subject:      Re: Memory issue?
> > > Date:         09/14/200
> > > Author:       Jim Reid <jim at rfc1035.com>
> > >
> > > If you're using BIND8, each zone needs a zoneinfo struct that occupies
> > > ~500 bytes (in round figures). On top of that memory is needed for the
> > > resource records in each zone.
> > >
> >
> > [snip]
> >
> > > Subject:      Re: BIND 8 memory leak symptoms
> > > Date:         11/19/2000
> > > Author:       Mark.Andrews <Mark.Andrews at nominum.com>
> > >
> > > It's unique recursive queries that make caches grow.
> > >
> > > Well every record has 32 bytes of overhead.  Every label has
> > > 20 bytes of overhead.  Each zone uses 324 bytes.  Then there
> > > are the hash tables.
> >
> > [snip]
> >
> > Do I apply same mem sizing for housekeeping to local (auth) RR as
> > cached RR? Currently I use the filesize to estimate mem requirement
> > for RR themselves, is that ok?  Not trying to split hairs, we have
> > one server that's on the skinny side.
> >
> > CT
> >
> >
> >
>         The sizes were taken from BIND 8.2.2-P7 on a BSDI 3.1 box.
>         The values are sizeof(struct databuf) - 4,
>         sizeof(struct namebuf) - 4 and sizeof(struct zoneinfo).
>
>         The "-4"s are because the end of the structure is dynamically
>         allocated array containing the rdata and label respectively.
>
>         Yes theses values apply regardless of whether it is authoritative
>         data or cached data.

Of course, if host-statistics are enabled, then this blows the estimates out
of the water.

Offhand, does anyone know if host-statistics add anything to
*authoritative* data?


- Kevin





More information about the bind-users mailing list