[rrd-developers] striping - initial observations

[kevin brintnall]

On Wed, Oct 22, 2008 at 01:54:48PM +0100, Daniel.Pocock at barclayscapital.com wrote:
> rrdcached is also in use:
>   /usr/bin/rrdcached -w 300 -z 300 -p /var/run/rrdcached/rrdcached.pid
> -l /var/run/rrdcached/rrdcached.sock
>
> - Using striping without rrdcached, the average IO request size becomes
> even bigger: from 8 sectors up to about 50 sectors.  I'm investigating
> why rrdcached reduces this figure.

Daniel,

Try removing the "-z" argument.  It causes RRD write times to be randomly
spread over the "-z" time.  This may reduce the degree to which your
writes are coalesced.

I'd be interested to see what your approach is for managing the striped
files once we need to rm individual RRDs.  In many cases, the set of RRDs
is a moving target.  It would also be nice if the "slots" in the stripe
files could be reused once a file was rm'ed..  Or re-packed when there are
a lot of holes.

Also, the striped RRDs must all have the same RRA/DS definitions right?
Then, it would make sense to set the stripe size to some multiple of a
full row..  For some K:

	strip_size = K * sizeof(rrd_value_t) * ds_cnt

-- 
 kevin brintnall =~ /kbrint at rufus.net/