[DRBD-user] Why is write performance so low?

Holger Kiehl Holger.Kiehl at dwd.de
Thu May 7 23:24:53 CEST 2009

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


On Thu, 7 May 2009, Lars Ellenberg wrote:

> On Thu, May 07, 2009 at 08:13:55PM +0000, Holger Kiehl wrote:
>>> local: 160 MB/s
>>> drbd disconnected: 70 MB/s
>>> AL-transaction latency appears to be:  32142 micro seconds
>>> you may want to verify this using e.g.
>>> dd if=/dev/zero of=ll_dev bs=512 count=1000 oflag=direct
>>>
>>> there.  now, how long does a synchonous single sector write take on you
>>> lower level device?
>>>
>> dd if=/dev/zero of=/dev/md4 bs=512 count=1000 oflag=direct
>> 1000+0 records in
>> 1000+0 records out
>> 512000 bytes (512 kB) copied, 3.38987 s, 151 kB/s
>
> right. though, your meta data is "internal",
> it is at the very end of the device,
> so all the disks in the md4 raid have to seek to the very end,
> possibly adjusting rotation speed, do that single sector write,
> then seek back (again possibly adjusting rotational speed),
> then write the next 4MB, seek back to the end, do the next transaction,
> seek to the front again.
>
> just for the fun of it, and to rule it out as possible cause,
> as an experiment, you could place DRBD meta data on a ram disk...
>
I will try this tommorrow and report what I get. Is there some documentation
on how to do this? How do I get the size of DRBD meta data?

Many thanks for all the help!

Holger



More information about the drbd-user mailing list