| [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] |
The tests I'm running are close to the typical I/O we see in our normal
production environment, ie seismic data processing. If I get time I'll
look at downloading some of the more usual benchmarking tools but I
can't promise anything.
In the meantime I've added a gigE switch and another client. Here are
the results for concurrent access to the ogfs partition over HyperSCSI.
C1 C2 MB/Sec
sequential write: 18 25
sequential read: 31 32
random rewrite: 5 4
Note, filesystem mounted noatime, and locking over separate 100Mbit
Ethernet.
Just to make my intentions clear, when I said I hope to have up to 32
clients, I will also have at least 6 more storage servers with more than
30TB of disk. We also plan to upgrade our current network, which is
2Gbit Myrinet and not gigE (although I'm using gigE in these tests) to
10GBit Infiniband - so I hope to have enough bandwidth.
Keep up the good work.
Thanks.
On Tue, 2003-12-16 at 14:50, Brian Jackson wrote:
> On Monday 15 December 2003 08:53 pm, Franco Broi wrote:
> > Just did some quick benchmarks of OGFS 0.3 vs ext3.
> >
> > Storage server: dual Athlon 2200+, 1GB memory, 5TB IFT IDE RAID
> >
> > Client machine: dual 2.4GHz Xeon, 3GB memory
> >
> > Network: 1Gbit Ethernet point-to-point between storage server and
> > client.
> >
> > All the tests use 8K block sizes and a 7GB file size. All results are in
> > MB/Sec. The filesystem size was 2TB. For ogfs I used LVM to create the
> > partitions for ci and data. I created a separate journal partition but
> > didn't use it.
> >
> > NBD = Network Block Device
> > HS = HyperSCSI
> >
> > local ext3 NFS NBD/ext3 HS/ext3 HS/ogfs
> > sequential write: 73 41 58 60 53
> > sequential read: 88 55 43 42 41
> > random rewrite: 29 11 15 17 14
>
> this is interesting, I would have thought ogfs would have been further behind,
> if you got some spare time, I'd be interested to see some small file
> benchmarks (bonnie++, dbench, etc.)
>
> >
> > This is the filesystem as mounted on the client machine.
> >
> > [franco@charlie6]$ time df /data22
> > Filesystem 1K-blocks Used Available Use% Mounted on
> > /dev/data22/data22 2042897724 15640632 2027257092 1% /data22
> > 0.000u 5.990s 0:06.00 99.8% 0+0k 0+0io 135pf+0w
> >
> > Some observations while configuring ogfs.
> >
> > The journal config doesn't appear to be optional, as stated in the help.
> > I just wanted to use n internal journals but had to list them all in the
> > config file. I might be pushing the limits here, but eventually I intend
> > to have up to 32 clients.
>
> see my other mail, but I do think it's a bug worth fixing if that is broken
>
> >
> > How does one use external journals for multiple clients? Do you have to
> > configure a separate journal device per client?
>
> also see my other email
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by: SF.net Giveback Program.
> Does SourceForge.net help you be more productive? Does it
> help you create better code? SHARE THE LOVE, and help us help
> YOU! Click Here: http://sourceforge.net/donate/
> _______________________________________________
> Opengfs-users mailing list
> Opengfs-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/opengfs-users
-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/
_______________________________________________
Opengfs-users mailing list
Opengfs-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opengfs-users