Re: [Bacula-users] client/server passwords

> I'd say your initial tls overhead was due to slow cipher

it's very strange as I have a fastest CPU on both of sides

storage: Intel Xeon CPU E3-1245 V2 @ 3.40GHz
client: Intel Core i7-4770 CPU @ 3.40GHz

And as I can see in the top output the CPU is not bottleneck at all, imho

On the storage I'm using the following filesystem

/dev/sda5 on /data type xfs (rw,relatime,attr2,inode64,noquota)

# cd /data
# dd if=/dev/zero of=tmp.file bs=8192 count=1000000 oflag=direct
1000000+0 records in
1000000+0 records out
8192000000 bytes (8.2 GB) copied, 65.5126 s, 125 MB/s

On Tue, May 12, 2015 at 10:24 PM, Dimitri Maziuk <dmaziuk AT bmrb.wisc DOT edu> wrote:

On 05/12/2015 01:50 PM, Alex Domoradov wrote:
>> Where's 23MB/s come from?
> from bacula-web - http://i.imgur.com/pEQwCvI.png
>
>> I get 200+MB/s on disk writes to raidz zfs backed by cheap "desktop"
> drives
> It would depend on "type" of files as I understood. I have a lot (~11M) of
> small files 20-50 Kbyte. So I don't believe that you can get 200 Mb/s on
> such small files ;)

Yeah, OK. I see thing like

> 1-May 20:05 starfish-sd JobId 7: Committing spooled data to Volume "BMRB-0001". Despooling 54,630,677 bytes ...
> 11-May 20:05 starfish-sd JobId 7: Despooling elapsed time = 00:00:01, Transfer rate = 54.63 M Bytes/second
> 11-May 20:05 starfish-sd JobId 7: Elapsed time=00:00:08, Transfer rate=6.791 M Bytes/second

in the log and I've no idea what any of it means: the spool is a
mirrored ssd so "6.791 M Bytes/second" is not the disk write speed. The
"54.63 M Bytes/second" just means "I despooled everything I had in a
second or less". On large sets where it despools the entire 50GB file
volume, reported ssd to spinning rust speed is

> 12-May 00:29 starfish-sd JobId 13: Writing spooled data to Volume. Despooling 49,807,364,491 bytes ...
> 12-May 00:31 starfish-sd JobId 13: End of medium on Volume "BMRB-0002" Bytes=49,807,326,888 Blocks=772,063 at 12-May-2015 00:31.
> 12-May 00:31 starfish-sd JobId 13: Despooling elapsed time = 00:02:26, Transfer rate = 341.1 M Bytes/second

-- that does sound reasonable.

Anyway, I'd say your initial tls overhead was due to slow cipher and you
shouldn't have a problem setting things up so that encryption overhead
is masked out by the storage write speed or the rate the clients can
produce the data or whatever the current bottleneck is.

--
Dimitri Maziuk
Programmer/sysadmin
BioMagResBank, UW-Madison -- http://www.bmrb.wisc.edu

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud 
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y

_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users