[rrd-developers] Re: How to get the most performance when using lots of RRD files

[Ole Bjørn Hessen]

One solution is to use more disk spindles - or spread writes on more
than one writer machine.

Another solution is to make the RRD-files smaller :-)

Another solution is to have more targets/counters in each RRD-file so
that each RRD-file update is really a set of updates. A large
sequential write on disk is much more easier on a disk than lots of
small random writes.

-

We use a NetApp / fileserver with lots of fast disk spindles.  Please
keep in mind that an RRD-update consists of reading and updating a
small header part at the start of the file, and a small random write
into the rest of the file.

The performance of the NetApp will probably be determined by how many
spindles you have, and how much of the header part of the RRD-file will
be in RAM cache for the reading of header and in a somewhat less degree
how much flash you have for caching updates before flushing updates to
disk spindles.

-

Beware that flash-disks are get bigger, cheaper and faster. That's the
future for RRD.

Ole Bjørn Hessen,
NMS-IP, PF-Nett, Telenor Networks

--
Unsubscribe mailto:rrd-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-developers-request at list.ee.ethz.ch?subject=help
Archive     http://lists.ee.ethz.ch/rrd-developers
WebAdmin    http://lists.ee.ethz.ch/lsg2.cgi