[DRBD-user] DRBD Performance

Lars Ellenberg Lars.Ellenberg at linbit.com
Fri Nov 17 14:21:41 CET 2006

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

/ 2006-11-17 12:25:02 +0100
\ Ralf Schenk:
> Hello !
> I've sucsessfully set up DRBD 8.0pre6 on 2 Ubuntu Server and XEN based
> Xeon Machines.
> I did some benchmarks and I'm not fully content with the performance.
> The machines are connected crossover via Fast Gigabit Adapters (PCI-X)
> with Jumboframes (MTU 9000) and netio shows near wirespeed throupout of
> about 116000 KByte/s in each direction.

jumbo frames not necessarily increase drbd throughput.
benchmark this.

> While benchmarking I saw that there was less data transferred over the
> wire than the benchmark had to write. Is there any compression used in
> the network protocol of drbd ?

> Is there best practice setting of
> max-buffers
> max-epoch-size
> sndbuf-size
> or other settings for a fast Gigabit interconnect and Protocol C usage ?

they are more dependant on the underlying storage device
than on the network link.
there are no "best practices", each setup behaves differently.

> At the moment performance decrease in writing speed of underlying
> device/drbd device is about 25%. Is that ok ?

whether it is "ok", you have to decide.
we have seen in our test setups less than 2%,
and expect this to become better.

: Lars Ellenberg                                  Tel +43-1-8178292-0  :
: LINBIT Information Technologies GmbH            Fax +43-1-8178292-82 :
: Schoenbrunner Str. 244, A-1120 Vienna/Europe   http://www.linbit.com :
please use the "List-Reply" function of your email client.

More information about the drbd-user mailing list