May 12, 201016 yr I am limited to 8MB/s write speed to my unraid from my w2k server, vista, and Win7 64bit machines. When doing parity checks it starts at 50MB/s and peeks to 80MB/s. I have a Dell Powerconnect 2724 gigabit switch and 5e wiring. My read speed from the unraid are about 45MB/s max. I have tried multiple unraid drives to multiple win7 drives and it is still sums up to 45MB/s. I doesn't matter how many drives it only uses about 40% gigabit speed. I can copy off my w2k server to my win7 and get 70% gigabit using multiple drives to multiple drives. So I don't think it is the switch. My unraid server is a Gigabyte MA-770-UD3 motherboard with a underclocked AMD Athlon Dual Core Processor 4450e stepping 02. My write speeds are still 8MB/s whether the CPU is running at 1Ghz or 2.3Ghz. I have 2Gigs of DDR2 800 and a pci ATI rageXL video card. In bios I have AHCI set for all sata ports. I have set my sata ports to IDE and things get worse. Writes were less than 8MB/s and parity checks were only 25MB/s so I have left it AHCI on. On or off, NCQ doesn't seem to change anything. My drives are as follows parity WDC_WD20EADS-11R_WD-WCAVY1156698 disk1 WDC_WD7500AACS-0_WD-WCASN0091321 disk2 WDC_WD20EADS-00R_WD-WCAVY1818092 disk3 WDC_WD20EADS-00S_WD-WCAVY0422389 disk4 WDC_WD20EADS-00R_WD-WCAVY2125130 disk5 WDC_WD20EADS-00R_WD-WCAVY1832813 Anyway I have search forums and found nothing. Before creating the parity drive I could have sworn I was copying to unraid server at 40MB/s. Otherwise I would still be copying to it today. Looking for suggestions. I have attached a syslog also. John MOBO link http://www.newegg.com/Product/Product.aspx?Item=N82E16813128376 syslog-2010-05-12.txt
May 12, 201016 yr First step, determine the speed of the lan connection by typing: ethtool eth0 then look for errors/collisions ifconfig eth0
May 12, 201016 yr Author Linux 2.6.32.9-unRAID. root@bNAS:~# ethtool eth0 Settings for eth0: Supported ports: [ TP MII ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Half 1000baseT/Full Supports auto-negotiation: Yes Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Half 1000baseT/Full Advertised auto-negotiation: Yes Speed: 1000Mb/s Duplex: Full Port: MII PHYAD: 0 Transceiver: internal Auto-negotiation: on Supports Wake-on: pumbg Wake-on: g Current message level: 0x00000033 (51) Link detected: yes root@bNAS:~# ifconfig eth0 eth0 Link encap:Ethernet HWaddr 6c:f0:49:73:de:15 inet addr:192.168.1.22 Bcast:192.168.1.255 Mask:255.255.255.0 UP BROADCAST NOTRAILERS RUNNING MULTICAST MTU:1500 Metric:1 RX packets:1131479 errors:0 dropped:0 overruns:0 frame:0 TX packets:352617 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:1586457766 (1.4 GiB) TX bytes:46524795 (44.3 MiB) Interrupt:26 Base address:0xe000 I think all look good? John
May 12, 201016 yr I don't see anything wrong either, but I figured it was worth a check first. If the LAN was connecting at 100Mbit/s it would have explained a lot.
May 12, 201016 yr If you got 40MB/s w/o parity, I would expect you would drop to about 10-12 MB/s with parity. First, do some write tests using dd and not across the wire.... that can tell you a lot. Second, try upgrading to 4.5.3
May 12, 201016 yr Not to thread hijack, but is that the rule of thumb? About 1/4 the write performance once you've added your parity disk? I have been copying without parity for speed, (and since it's a copy, losing it doesn't matter) and have been seeing 49MB/s write rates pretty consistently. So after adding parity, around 12MB/s should be expected?
May 12, 201016 yr Author 2TB WD green root@bNAS:~# dd if=/dev/sdd of=/dev/null count=8192000 8192000+0 records in 8192000+0 records out 4194304000 bytes (4.2 GB) copied, 67.3944 s, 62.2 MB/s 750GB WD green root@bNAS:~# dd if=/dev/sdb of=/dev/null count=8192000 8192000+0 records in 8192000+0 records out 4194304000 bytes (4.2 GB) copied, 87.7501 s, 47.8 MB/s And I'm running 4.5.3
May 12, 201016 yr Author On a side note the preclear times are 30 hours for the 2TB drives and they write at about 100MB/s and read at 85MB/s
May 12, 201016 yr Not to thread hijack, but is that the rule of thumb? About 1/4 the write performance once you've added your parity disk? I have been copying without parity for speed, (and since it's a copy, losing it doesn't matter) and have been seeing 49MB/s write rates pretty consistently. So after adding parity, around 12MB/s should be expected? Baring a bottleneck on your motherboard BUS and disk controllers, the rate is entirely dependent upon the rotational speed of the slowest disk involved. I don't think the benchmark given is true. Faster rotating disks will allow you to write to the array faster. See this post, and the thread it links to: http://lime-technology.com/forum/index.php?topic=5676.msg53165#msg53165
May 12, 201016 yr Not to thread hijack, but is that the rule of thumb? About 1/4 the write performance once you've added your parity disk? I have been copying without parity for speed, (and since it's a copy, losing it doesn't matter) and have been seeing 49MB/s write rates pretty consistently. So after adding parity, around 12MB/s should be expected? I would expect it to be higher then 12MB/s. How much ram is in your system? I used to get around 12-15MB/s before unRAID 4.5. Now it's much higher. Somewhere between 23-32MB/s using teracopy from a local disk to a remote unRAID disk share (not the user share which adds overhead). It also depends on how full your disk is. The speed of your parity drive is also a factor. I am using 1.5TB 7200RPM drive for Parity.
May 12, 201016 yr Author Read root@bNAS:~# dd if="/mnt/disk2/Acronis/Backup/XPutil.tib" of=/dev/null 7837103+1 records in 7837103+1 records out 4012597006 bytes (4.0 GB) copied, 58.1763 s, 69.0 MB/s Write root@bNAS:~# dd if="/mnt/disk2/Acronis/Backup/XPutil.tib" of=/mnt/disk4/Movies/X P.tib 7837103+1 records in 7837103+1 records out 4012597006 bytes (4.0 GB) copied, 505.146 s, 7.9 MB/s Some more dd stuff. It's starting to look like a Mobo problem or cabling?
May 12, 201016 yr do dd if=/dev/zero of=//mnt/disk4/test.dd count=8192000 do /root/mdcmd status | strings | grep rdevName then choose a drive hdparm -tT /dev/sd? where ? is the letter of your drive. Where ? = the drive letter of one of your rdevName drives.
May 12, 201016 yr If it is only disk4 that is slow, it might be the culprit. If writing to a different drive is also slow, it might be the parity disk. Joe L.
May 12, 201016 yr Author root@bNAS:~# dd if=/dev/zero of=//mnt/disk4/test.dd count=8192000 8192000+0 records in 8192000+0 records out 4194304000 bytes (4.2 GB) copied, 560.928 s, 7.5 MB/s root@bNAS:~# dd if=/dev/zero of=//mnt/disk5/test.dd count=8192000 8192000+0 records in 8192000+0 records out 4194304000 bytes (4.2 GB) copied, 537.781 s, 7.8 MB/s Linux 2.6.32.9-unRAID. root@bNAS:~# /root/mdcmd status | strings | grep rdevName rdevName.0=sda rdevName.1=sdb rdevName.2=sdc rdevName.3=sdd rdevName.4=sde rdevName.5=sdf root@bNAS:~# hdparm -tT /dev/sdf /dev/sdf: Timing cached reads: 1122 MB in 2.00 seconds = 560.77 MB/sec Timing buffered disk reads: 308 MB in 3.02 seconds = 101.97 MB/sec root@bNAS:~# hdparm -tT /dev/sda /dev/sda: Timing cached reads: 1000 MB in 2.00 seconds = 500.06 MB/sec Timing buffered disk reads: 286 MB in 3.01 seconds = 94.97 MB/sec Not sure what all the results are telling me?
May 12, 201016 yr Not sure what all the results are telling me? The hdparm results show that the drives themselves can communicate at adequate speed. Why there is such a diminished throughput when parity writes occur is beyond me at the moment.
May 12, 201016 yr Do you have data on all the drives? Can you swap the parity drive with another WD20EADS temporarily?
May 12, 201016 yr I just switched from a WD15EADS parity drive to a WD20EARS parity drive. This took me about 30 hours to preclear, so about the same as you saw. When I build parity on it I get around 77MB/s speed, so about the same as you. When I am copying large files to the parity-protected array I get about 20-24MB/s (I'm copying over gigabit ethernet). My array is composed of 8 SATA drives, the 2TB unit, 5 x WD15EADS (1.5TB drives) and 2 x WD10EADS (1TB drives) - all of these are WD green drives, so I would expect your system to get at least 20MB/s. Have you done SMART reports to see if any of the drives are unhappy? Regards, Stephen
May 12, 201016 yr Author I just got finished building another unraid server with a 1.5TB Seagate 7200rpm, 500GB Seagate 7200rpm, and a 500GB Samsung 7200rpm. It is a Nforce4 board with a AMD 4000+ ,1 gig of ram and onboard gigabit. Parity check was at 60MB/s for the first 1/3 and over 100MB/s for the last 2/3 of the check. Anyway once parity check was done, I was able to copy a 8GB MKV to it at 40MB/s. I was able to read off the unraid at 55MB/s. Well now I know that it is not my network or server. I think I will just have to swap the Gigabyte GA-MA770-UD3 motherboard and see what results I get. I have smart checked all the drives and the WDC_WD7500AACS is the only one that had logged some activity. I'll attach the reports for the 750GB and one of the 2TB. All the 2TB look about the same. smartctl.txt
May 13, 201016 yr Author Changed out my GA-MA770-UD3 for a GA-MA590-S5. Still getting Less then 10MB/s. Still in AHCI and all stock bios settings. I'm not even underclocking the CPU. I even changed all the sata cable. I think I'm going to throw in the white flag. I guess 5400rpm drives are just that slow. When I change out the 750GB with a 2TB drive I'll change the parity first. Just to make sure it isn't the parity drive causing problems.
May 13, 201016 yr No, don't give in. You are seeing atypical results for 5400 rpm drives. My array has faster parity-protected writes than you're seeing and I'm using nothing but 5400 rpm drives. If I'm reading correctly, he's doesn't have any EARS and is using EADS drives which do not have the jumper.
May 13, 201016 yr FWIW, here's the results on my 2TB EADS system while cache_dirs is running too: # df Filesystem 1K-blocks Used Available Use% Mounted on /dev/md1 1953454928 1086196624 867258304 56% /mnt/disk1 /dev/md2 1953454928 810735724 1142719204 42% /mnt/disk2 # dd if=/dev/zero of=/mnt/disk1/test.dd count=8192000 8192000+0 records in 8192000+0 records out 4194304000 bytes (4.2 GB) copied, 148.896 s, 28.2 MB/s # dd if=/dev/zero of=/mnt/disk2/test.dd count=8192000 8192000+0 records in 8192000+0 records out 4194304000 bytes (4.2 GB) copied, 124.539 s, 33.7 MB/s # /root/mdcmd status | strings | grep rdevName rdevName.0=sdb rdevName.1=sdc rdevName.2=sdd # hdparm -tT /dev/sdd /dev/sdd: Timing cached reads: 3702 MB in 2.00 seconds = 1851.81 MB/sec Timing buffered disk reads: 290 MB in 3.02 seconds = 96.08 MB/sec # hdparm -tT /dev/sdb /dev/sdb: Timing cached reads: 3784 MB in 2.00 seconds = 1892.68 MB/sec Timing buffered disk reads: 270 MB in 3.00 seconds = 89.86 MB/sec
May 14, 201016 yr Sorry to Hijack, but I'm getting terrible speeds as well. I saw Joe's post where he says: I don't see anything wrong either, but I figured it was worth a check first. If the LAN was connecting at 100Mbit/s it would have explained a lot. I ran the "ethtool eth0" command and here is my output: Settings for eth0: Supported ports: [ TP ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Supports auto-negotiation: Yes Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Advertised auto-negotiation: Yes Speed: 100Mb/s Duplex: Full Port: Twisted Pair PHYAD: 0 Transceiver: internal Auto-negotiation: on Supports Wake-on: pg Wake-on: d Current message level: 0x00000000 (0) Link detected: yes How do I get my server to run at 1000Mb/s and would this explain my slow speeds? Thanks, Kent
May 14, 201016 yr How do I get my server to run at 1000Mb/s 1. The device you're connecting it to must also be able to do 1000Mb/s. 2. You must use a good cable, CAT5e or CAT6. Anything less than that and they'll negotiate slower speed.
May 14, 201016 yr Purko, Thanks for the response. I'm connecting to a gigabit switch, but I will have to check my cables. I also noticed that when copying a file form one drive to another using midnight commander my speed never gets over 10 MB/s. I wonder if I have bigger problems with my machine. Any ideas what I should look at ? Thanks again Kent
Archived
This topic is now archived and is closed to further replies.