Re: 4x lower IOPS: Linux MD vs indiv. devices - why?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Andrey,

Changing sector to 4k is easy, this can really help. see DCT manual, it’s there.
This can be relevant for you https://itpeernetwork.intel.com/how-to-configure-oracle-redo-on-the-intel-pcie-ssd-dc-p3700/

After overcoming my issues with isdct, and reformatting the NVMes to 4k sector size, success!

9.5 mio IOPS =)

This is another 34% faster than before.

So: thanks a bunch for your tip!

Cheers,
/Tobias


Next steps:

- approach MD developers about bottlenecks there
- approach PostgreSQL about using pread/pwrite (instead of lseek/read/write)


randread-individual-nvmes: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=128
...
fio-2.1.11
Starting 128 threads
Jobs: 128 (f=2048): [r(128)] [100.0% done] [37244MB/0KB/0KB /s] [9534K/0/0 iops] [eta 00m:00s] randread-individual-nvmes: (groupid=0, jobs=128): err= 0: pid=25406: Tue Jan 24 15:57:19 2017
  read : io=1083.9GB, bw=36964MB/s, iops=9462.8K, runt= 30026msec
  cpu          : usr=9.00%, sys=77.01%, ctx=49252920, majf=0, minf=16512

--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux Kernel]     [Linux SCSI]     [Linux IDE]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux SCSI]

  Powered by Linux