February 8, 201214 yr I have built a new box with unRaid Pro 5.0b14. I am experiencing mediocre read speeds when reading from harddisk through samba share, but I get fast reading when reading from a ssd drive or when reading the hdd-file a second time after the file is cached in memory. The read speed of the file is fast when reading file using dd through terminal. I use Windows 7 as client. Harddisk performance Direct read through dd >dd if=/mnt/disk1/storage/temp/Video/Minisekterne.dvr-ms of=/dev/null count=1000000 626207+1 records in 626207+1 records out 320618435 bytes (321 MB) copied, 3.99904 s, 80.2 MB/s dd if=/mnt/user/media/Film/Huge8GbMovie.iso of=/dev/null count=10000000 10000000+0 records in 10000000+0 records out 5120000000 bytes (5.1 GB) copied, 61.6959 s, 83.0 MB/s Copy file through samba user share or directly from /mnt/disk1 gives 44MB/s Copying the same file again gives 105MB/s (now the file is cached in memory) The samba performance is independent of whether I access through \\tower\disk1\myshare or \\tower\myshare\ From SSD (in array, though I get same results when ssd is mounted as a cache drive) Copy file via samba gives 88MB/s-98MB/s Copy same file again via samba gives 106 MB/s I'm copying to an ssd on the windows machine, so the windows drive is not a bottleneck. Using bwm-ng diskmonitoring and network monitoring in the background I can verify that nothing is running on the disk when I'm not copying. I looked at non-averaged 0.5s bwg-ng stats while copying the 5GB. Reading with dd, it read with 105MB/s most of the time, but dropped occasionally to 10MB/s (hence the 80MB/s avg), copying file with samba I got 44MB/s. bwm-ng showed me typically 40-50MB/s, dropping ocassioully to 10MB/s (fragmentation?). It's a brand new machine, which has been filled with data in mostly one big copy operation, so it should be pretty defragmented. What ever scheduler and read-ahead I activate in unMenu, I get the same results. The above where results were with /sys/block/sda/queue/scheduler [noop] deadline cfq /sys/block/sdb/queue/scheduler [noop] deadline cfq /sys/block/sdc/queue/scheduler [noop] deadline cfq /sys/block/sdd/queue/scheduler [noop] deadline cfq /dev/md1 1024 /dev/md2 1024 I have attached a syslog. I have the following hardware MB: Asus PcB WS CPU: Intel xeon 1225 Lan: 2 x Intel82574L GbE Lan (only one connected at the moment) Parity: 3TB WD30EZRX 5400RPM (connected to mb 6GB/s port) Data1: 2TB SAMSUNG_HD204UI 5400RPM (connected to mb 3GB/s port) Data2: 120GB OCZ Vertex 3 (for test purposes ) (connected to mb 6GB/s port) No pci or pci-e sata card I have read a few recent post from people in the forums reporting 90+MB/s transfers. Though that might not have been through samba but apple client. Is this the best to be hoped for with unRaid samba or is something wrong? Any thoughts appreciated Alex syslog-2012-02-08.txt
Archived
This topic is now archived and is closed to further replies.