[OpenIndiana-discuss] ZFS remote receive

Timothy Coalson tsc5yc at mst.edu
Thu Nov 1 03:37:51 UTC 2012


On Wed, Oct 31, 2012 at 8:44 PM, Richard Elling <
richard.elling at richardelling.com> wrote:

> > > On the target system I am seeing writes up to
> >> 160 MB/s with frequent zpool iostat probes. When iostat probes are up to
> >> 5s+, there is a steady stream of 62 MB/s.
> >
> > I believe this *may* mean that your networking buffer receives data
> > into memory (ZFS cache) at 62Mb/s, then every 5s the dirty cache
> > is sent to disks during TXG commit at whatever speed in can burst
> > (160Mb/s in your case).
>
> More likely: straight pipe send | receive is a blocking configuration. This
> is why most people who go for high speed send | receive use a buffer,
> such as mbuffer, to smooth out the performance. Check the archives,
> this has been rehashed hundreds of times on these aliases.


Thank you very much for rehashing it again, I stuck "| mbuffer -b 8192 -m
256M -q 2> /dev/null |" (some preliminary testing seemed to indicate it
wanted 8192 blocksize for pipes, and when run from cron it produces an odd
warning message) in the middle of my send | ssh recv pipe and was rewarded
with this over gigabit ethernet:

received 8.63GB stream in 89 seconds (99.3MB/sec)

Previously I was getting 70MB/s or less, even after switching to arcfour128
for ssh cipher.  My only gripe is that mbuffer doesn't have a manpage on
OpenIndiana.

Tim


More information about the OpenIndiana-discuss mailing list