slower performance than expected (12MB/sec)

[evans036]

i am a noob to unraid and just LOVE it so far.

 

i get read performance of about 12MB/sec from Unraid to windows boxes (using 1GB file).

 

verified all NICs running at 1Gb, full duplex

 

changed NIC and got same performance

 

within Unraid server command:

cp /mnt/disk1/bigfile /dev/null

 

runs at or above 60MB/sec (depending on the disk that i use)

 

during tests Unraid is otherwise idle, CPU does not go above 20%

 

also: all transfers to and from the unraid share seem to run at about 12MB/sec.

 

i expect slower writes, but not reads

 

oh nearly forgot: version is Unraid 4.7

 

anyone got any clues? i sure dont.

 

thanks a lot in advance.

 

steve

[Johnm]

perhaps you should post a syslog and list of your network equipment.

 

11MB/s(ish) is the upper limit of a 100Mb switch/hub/router.

It sounds like you are bottlenecked somewhere without seeing your info.

if i had to guess, you have a non-gigabit device in your path from desktop to unRAID...

[evans036]

john,

 

that was my thought too but...

 

12MB/sec is faster than 100Mb (although you are right it IS close). btw, i have 3 hosts on this one switch: Unraid, Win 7 & Win xp. The Win 7 & Win XP transfer between each other at around 22MB/sec. so that traffic is going thru the same switch and that is definitely > 100Mbs.

 

as requested, my syslog is attached.

 

btw, Unraid server version is 4.7

 

thanks a lot for your input,

 

steve

 

 

syslog.zip

[evans036]

looking at this a little more closely...

 

while 'top' shows cpu < 20%, vmstat is showing cpu around 70%

 

why the discrepancy? i think this might be a dual core processor - maybe thats confusing these tools?

 

also, io waits are about zero so its not the disk.

 

and of its not disk, it must be cpu or network

 

i also note that samba is taking up most of the cpu, followed by dhcpd - why is dhcpd running? isnt that for a dhcp server?

 

btw, this is also an old processor (i think i got it in 2005) and even when i bought it it was not high end.

 

so if i am too cheap to buy newer hardware, i might just have to live with these speeds.

 

steve

[Darts]

Can you try to install a cache drive and see if this helps?

[ixnu]

Can you try it the other way?

 

Try copying something from your other machines to the unraid box as a speed test. Let me know if you need examples.

[dgaschk]

The syslog is full of these:

 

Dec 18 10:08:49 BIGASS unmenu[1274]: bad method -    51301  5693423 45958680  1942900    45817  5274694 42564088  3897330        0  1466640  584017051301-^M
Dec 18 10:11:54 BIGASS last message repeated 2 times
Dec 18 10:13:57 BIGASS unmenu[1274]: bad method -    51924  5767188 46553784  1970770    46460  5348431 43153032  3946740        6  1486430  591763051924-^M
Dec 18 10:16:01 BIGASS unmenu[1274]: bad method -    54473  6073529 49024912  2076410    49149  5654626 45627168  4169380        3  1571380  624587054473-^M
Dec 18 10:19:07 BIGASS unmenu[1274]: bad method -    58573  6541300 52793792  2243070    53282  6121447 49397832  4494110        6  1698640  673746058573-^M
Dec 18 10:20:15 BIGASS unmenu[1274]: bad method -    59105  6602274 53291936  2265950    53821  6183175 49895968  4540620        0  1715150  680651059105-^M
Dec 18 10:22:12 BIGASS unmenu[1274]: bad method -    60855  6815015 55007864  2338400    55632  6395842 51605696  4681630        6  1770830  702018060855-^M
Dec 18 10:23:14 BIGASS unmenu[1274]: bad method -    62196  6976050 56306872  2399540    57002  6556844 52905688  4794400        5  1813760  719428062196-^M

 

See here: http://lime-technology.com/forum/index.php?topic=5568.msg88221#msg88221

 

Once the "bad method" messages are gone collect a new syslog after running for at least an 30 min and performing a long transfer.