Encrypted 2-drive mirrored RAID setup

[mastashake57]

iperf tests...

$this->bbcode_second_pass_code('', '[root@pogoweb ~]# iperf -s
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[ 4] local 192.168.48.20 port 5001 connected with 192.168.48.10 port 44029
[ ID] Interval Transfer Bandwidth
[ 4] 0.0-10.0 sec 572 MBytes 479 Mbits/sec
[ 5] local 192.168.48.20 port 5001 connected with 192.168.48.10 port 44030
[ 5] 0.0-10.0 sec 648 MBytes 542 Mbits/sec
[ 4] local 192.168.48.20 port 5001 connected with 192.168.48.10 port 44031
[ 4] 0.0-10.0 sec 672 MBytes 564 Mbits/sec')

[sambul13]

Interesting. Try converting Samba speeds to MB/s for easier comparison. The RAID enclosure is hooked to Pogo via USB.

Would the results change if you format your non-boot /dev/sda3 partition to EXT4 or XFS? Most should be much higher. Try testing your non-boot partition anyway - it should be faster than writing to sda1.

The above are speed results for the 1st drive - I presume they are identical for the 2nd drive? How about proc load as shown by Htop at writing to both RAID drives simultaneously? How it depends on FS type and source (LAN vs Local) - again pls try with EXT4 or XFS, since EXT3 is slower?

[mastashake57]

Ok, let me see what I can do. In the meantime, here's iperf with the -f M flag, 3 separate client runs:

$this->bbcode_second_pass_code('', '[root@pogoweb ~]# iperf -s -f M
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 0.08 MByte (default)
------------------------------------------------------------
[ 4] local 192.168.48.20 port 5001 connected with 192.168.48.10 port 44033
[ ID] Interval Transfer Bandwidth
[ 4] 0.0-10.0 sec 671 MBytes 67.0 MBytes/sec
[ 5] local 192.168.48.20 port 5001 connected with 192.168.48.10 port 44034
[ 5] 0.0-10.0 sec 679 MBytes 67.9 MBytes/sec
[ 4] local 192.168.48.20 port 5001 connected with 192.168.48.10 port 44035
[ 4] 0.0-10.0 sec 684 MBytes 68.4 MBytes/sec')

[sambul13]

Something must be wrong with these results. Max USB 2.0 throughput is 35MB/s. So, what exactly these results reflect - copy from PC to Pogo RAM? That's assuming, there is no double traffic to feed 2 RAID drives simultaneously, because your enclosure has only one USB port.

[mastashake57]

Well, I thought you wanted to see throughput from 2 different devices via Network. This is a client <==> server network throughput test with iperf. Is there something else you had in mind?

[sambul13]

So the test seems to reflect read from & write to RAM (not to HD), when 2 Pogos are hooked via a Gigabyte router? Its interesting, but I wanted to find max Samba COPY speed from a Windows or Linux PC to your RAID drives in USB enclosure hooked to Pogo, since that's the task you'll be doing often. Your test seems to show that Samba is not a bottleneck here, while USB or Pogo proc capacity at write workload can be.

[WarheadsSE]

http://en.wikipedia.org/wiki/Universal_ ... us#USB_2.0

http://web.forret.com/tools/bandwidth.a ... +(Hi-Speed)+%5BUniversal+Serial+Bus%5D


Effective doesn't always mean actual, either.

[sambul13]

Yes, it seems maximum theoretical throughput for USB 2.0 is 60 MB/s, while effective throughput reflecting bandwidth utilization efficiency is maxed at 35 MB/s. I get actual 30 MB/s write speed over Samba to a single EXT4 drive in USB 2.0 enclosure hooked to Pogo. Interesting, how the RAID will perform.

[moonman]

USB 2.0 is half-duplex master-slave. So 480Mb/s / 2 = 240 Mb/s. There are 8 bits in a Byte so 240 / 8 = 30MB/s.

[WarheadsSE]

Thanks for pointing out the duplexity