Disk Speed

karlochacon · 03-14-2011, 04:54 PM

hi guys

I have a linux server which has about 5 volumes from SAN (fiber channel) now I need to measure the transfer rate between one LUN which is a Logical Volume to another LUN which is another Logical Volume.

so basically this server has 5 LUNs from SAN
each SAN volume is a logical volume presented to linux

any tool to measure transfers rate (Megabytes per second)

so while I copy a file like 1 Terabyte I can get MB/second

thanks a lot

macemoneta · 03-14-2011, 05:41 PM

I use dd_rescue for that. You can test with different block sizes as well.

For example:

Code:

$ dd_rescue -b 1M infile outfile

dd_rescue: (warning): Don't use sparse writes for non-seekable output
dd_rescue: (info) expect to copy 4572256kB from infile
dd_rescue: (info): ipos:   4571136.0k, opos:   4571136.0k, xferd:   4571136.0k
                   errs:      0, errxfer:         0.0k, succxfer:   4571136.0k
             +curr.rate:    47123kB/s, avg.rate:    56313kB/s, avg.load:  3.9%
             >----------------------------------------.<  99%  ETA:  0:00:00 
dd_rescue: (info): read infile (4572256.0k): EOF
dd_rescue: (info): Summary for infile -> outfile
dd_rescue: (info): ipos:   4572256.0k, opos:   4572256.0k, xferd:   4572256.0k
                   errs:      0, errxfer:         0.0k, succxfer:   4572256.0k
             +curr.rate:   114262kB/s, avg.rate:    56320kB/s, avg.load:  3.9%
             >-----------------------------------------< 100%  ETA:  0:00:00

karlochacon · 03-14-2011, 06:08 PM

looks I don't have dd_rescue

Quote:

Originally Posted by macemoneta

I use dd_rescue for that. You can test with different block sizes as well.

For example:

Code:

$ dd_rescue -b 1M infile outfile

dd_rescue: (warning): Don't use sparse writes for non-seekable output
dd_rescue: (info) expect to copy 4572256kB from infile
dd_rescue: (info): ipos:   4571136.0k, opos:   4571136.0k, xferd:   4571136.0k
                   errs:      0, errxfer:         0.0k, succxfer:   4571136.0k
             +curr.rate:    47123kB/s, avg.rate:    56313kB/s, avg.load:  3.9%
             >----------------------------------------.<  99%  ETA:  0:00:00 
dd_rescue: (info): read infile (4572256.0k): EOF
dd_rescue: (info): Summary for infile -> outfile
dd_rescue: (info): ipos:   4572256.0k, opos:   4572256.0k, xferd:   4572256.0k
                   errs:      0, errxfer:         0.0k, succxfer:   4572256.0k
             +curr.rate:   114262kB/s, avg.rate:    56320kB/s, avg.load:  3.9%
             >-----------------------------------------< 100%  ETA:  0:00:00

macemoneta · 03-14-2011, 06:16 PM

So, install it. You don't specify what distribution you're using, but on most rpm based it's:

Code:

yum install dd_rescue

On deb based distributions:

Code:

apt-get install ddrescue

slimm609 · 03-14-2011, 08:50 PM

hdparm -t -T /dev/sda or whatever the lun is mapped to

Quote:

/dev/sda1:
Timing cached reads: 2442 MB in 2.00 seconds = 1220.50 MB/sec
Timing buffered disk reads: 400 MB in 3.00 seconds = 133.15 MB/sec

macemoneta · 03-14-2011, 09:46 PM

The test done by 'hdparm -tT' is a short test, and doesn't provide sustained throughput information. For example, compare my 4GB transfer test above (56MB/s) with hdparm:

Code:

# hdparm -tT /dev/md1

/dev/md1:
 Timing cached reads:   4742 MB in  2.00 seconds = 2373.03 MB/sec
 Timing buffered disk reads: 388 MB in  3.00 seconds = 129.27 MB/sec

It's more than 100% off.

karlochacon · 03-15-2011, 09:05 AM

so looks like dd_rescue is the way to go

Quote:

Originally Posted by macemoneta

The test done by 'hdparm -tT' is a short test, and doesn't provide sustained throughput information. For example, compare my 4GB transfer test above (56MB/s) with hdparm:

Code:

# hdparm -tT /dev/md1

/dev/md1:
 Timing cached reads:   4742 MB in  2.00 seconds = 2373.03 MB/sec
 Timing buffered disk reads: 388 MB in  3.00 seconds = 129.27 MB/sec

It's more than 100% off.