April 9, 201115 yr Hi All, For a while now I have been getting good performance from my unRAID server (from a mac SMB ~60MB/s write and ~70MB/s read) and I have noticed recently that the write performance has come way down (~9MB/s write via SMB). I attached the shares via NFS and no change in write speed (from a mac NFS ~8MB/s write and ~95MB/s read). To isolate the mac out of the equation I ethernet (gigabit) connected a windows7 laptop and no change (from windows ~8MB/s write and ~70MB/s read). Using unmenu/mymain and selecting the "performance" link a snapshot shows the following parity 6.2MB/s read IO 6.2MB/s write IO disk2 6.2MB/s read IO 6.2MB/s write IO I am using a cd 650MB iso as a test. Getting it off the server is great putting it on the server is slow. Note1 when the transfer to the server starts for a second the transfer is great (nfs ~95MB/s) then drops down and bounces up and down giving the above write numbers. Note2 when I run a parity check its starts running at ~98MB/s (usually starts at 107MB/s) and then settles after a few minutes to ~105MB/s which is about right Note3 nothing else is happening with the server Note4 syslog looks the same as always (attached) When I am transferring the test file to the server and looking at performance on the unRAID server why is there equal amounts of read and write. I never noticed this before. Its like unRaid is depositing bits of data on the data disk + updating parity then reading it off the data drive and parity drive to verify??? Anybody have an idea what may be going on......I dont believe I have changed anything? Thanks syslog-2011-04-10.txt
July 4, 201115 yr Tagging along for some answers... I'm having the same issue. 80+ read / 3 +/- write (with or without a cache drive)
July 4, 201115 yr Unless I'm mistaken that performance view shows the net performance; if one of the disks were having write issues the other disk's writes would also look bad. Can we get a SMART report? (edit: for both drives) Any chance your network hardware is involved? ifconfig eth0 on unRAID, and then maybe power cycle the switch/router it's connected to.
July 4, 201115 yr When writing to a parity protected array unRAID must FIRST read the disk block being written from both the data disk and the parity disk, then calculate parity based on the existing and new contents being written, then write the new values to the disk. You'll always see an equal number of reads and writes. It is not what is causing your slowness, but the need for the disk to rotate between the read and subsequent write to get to the same sector is what limits "write" speeds on most systems to 35MB/s or so. The speed you are reporting seems more like a LAN connection running at 100Mb/s rather than 1000Mb/s.
Archived
This topic is now archived and is closed to further replies.