April 4, 200818 yr With a Cache Drive, I, in theory should see better write performance, correct? My first cache write attempt is a single 4.5GB DVD VIDEO_TS Folder to the MOVIES User share that has the Cache drive enabled. It is taking about 15 minutes (from a Vista SP1 PC) and is averaging between 5.5-6 MB/Sec(per Vista Windows Explorer). I can open the \CACHE Share and see that the folders were created and is it now copying to the Cache Drive and I think is the intention. The Drive itself is a 300 GB SATA Seagate (7200.10 I believe), connected via eSATA to my P5E-VM DO MB (AHCI). The 1.5Gps Jumper has been removed... Just was expecting a little faster from the Cache Drive thing, but maybe something else going on... SW2
April 4, 200818 yr Your xfer speed is WAY below disk IO capability... I believe you have a lan IO problem. What is your read speed form unRAID?
April 4, 200818 yr Definitely something else is not right. You should get double that just going to normal parity-protected disk. We get 45+MB/sec dragging a file via explorer from a (PC) local disk to the 'cache' share using XP.
April 4, 200818 yr Author ...but on a positive note, the Mover App did it's thing overnight... (no idea how fast that happened since I wasn't awake at 3-4am) Looking at my Syslog, it looks like the Link to my NIC dropped at one point and then went back on at 100 instead of 1000 (It was 1000 to begin with)... I guess I need to check my NIC Cable connection and such... SW2
April 4, 200818 yr Author What do y'all use to test your speeds? I was using DUMeter, that I have to assume is "close"... http://www.dumeter.com
April 4, 200818 yr What do y'all use to test your speeds? I don't like synthetic tests. I write or read an actual .iso file and calculate the transfer speed. Bill
April 4, 200818 yr Dumeter is fine, or even just plain old windows task manager. On the unRAID side, bwm-ng is good.... particularly when watching disk write performance. I agree with Bilped, with one caveat. A utility like Dumeter or task manager will show you spikes and other information that a "copy .iso and use stopwatch" method will not.
April 4, 200818 yr Author I rebooted and it's coming back 100 Mbps only....Maybe a tweaked cable or something... Apr 4 08:28:04 Media kernel: e1000: eth0: e1000_watchdog: NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX Apr 4 08:28:10 Media kernel: e1000: eth0: e1000_watchdog: NIC Link is Down Apr 4 08:28:13 Media kernel: e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps Full Duplex, Flow Control: RX/TX Apr 4 08:28:13 Media kernel: e1000: eth0: e1000_watchdog: 10/100 speed: disabling TSO
April 4, 200818 yr Hi all, when i tested unRAID under Vista (not SP1) the performance was bad (5/6 MB/s) and it was a bug, with SP1 it's corrected ? For you problem SuperW2, recently i had the same problem as you and after i changed the cable it's work fine : 1000Mbps.
April 4, 200818 yr I agree with Bilped, with one caveat. A utility like Dumeter or task manager will show you spikes and other information that a "copy .iso and use stopwatch" method will not. I also look at task manager, but it is a secondary concern. My primary concern is the user experience and that is measured in time. Backing into a rate allows for easier comparisons. Bill
April 5, 200818 yr Author OK, back on point, I've tried 3 different Cat5e and Cat6 Cables (and 2 different Gigabit Switches, all of which I have no issues getting 1000 speed on other systems) into my unRAID server and each time the Link starts up at 1000 and then drops and goes back down to 100. I'm thinking its not a physical cable issue at this point... Here a a full Syslog after a recent reboot (after the latest Cable/Swtich swap). Any other ideas? I might go back to a Beta 3 and see if I see a similar situation... *** EDIT *** Going back through all the working 4.3 Beta (1, 2, and 3) resulted in the same, Link 1000, Drop, then Link 100...Going all the way back to my original 4.2.1 does not exhibit the same thing (so far it links at 1000, does not drop, nor drops back down to 100)... So while I'm no expert, I'm thinking something with the 4.3 core that is causing this link speed thing to drop.
April 5, 200818 yr Except for a drive problem (see below), your syslog confirms your observation, nothing else indicated. I located some interesting posts online which *might* be helpful, at least give you something to try. Start with this thread http://lkml.org/lkml/2007/5/1/542 (same behavior as you), then proceed to this thread http://lkml.org/lkml/2007/5/4/488. Also see this thread http://www.mydatabasesupport.com/forums/solaris/334867-1000mbps-connection-problem-between-linux-solaris.html. There wasn't much else relevant online. The "management enable" setting in the BIOS is a possibility, easy enough to check and test. Your syslog does show problems with sdh, the Seagate 300GB. I suggest moving it to a different SATA controller. Does it 'seem' to be working correctly? I really don't like all of the errors associated with it.
April 5, 200818 yr Author Except for a drive problem (see below), your syslog confirms your observation, nothing else indicated. I located some interesting posts online which *might* be helpful, at least give you something to try. Start with this thread http://lkml.org/lkml/2007/5/1/542 (same behavior as you), then proceed to this thread http://lkml.org/lkml/2007/5/4/488. Also see this thread http://www.mydatabasesupport.com/forums/solaris/334867-1000mbps-connection-problem-between-linux-solaris.html. There wasn't much else relevant online. The "management enable" setting in the BIOS is a possibility, easy enough to check and test. Your syslog does show problems with sdh, the Seagate 300GB. I suggest moving it to a different SATA controller. Does it 'seem' to be working correctly? I really don't like all of the errors associated with it. Thanks RobJ...I'll take a look at those links this evening...I'm looking, but not even finding a BIOS setting for Management Enable on my P5B-VM DO The only 300 GB Seagate on the system is the brand new one that I was adding as the Cache Drive with Beta 5...it's connected via eSATA, but does "appear" to be acting normal, but I only plugged it in the other day, so it's hard to say. I could swap the eSATA cable on it too as I have a few laying around. *** EDIT *** This will probably show my ignorance with Linux, but is there a way to change or update which NIC driver is being used? It seem this kernel is using version "7.3.20-k2-NAPI" for the e1000 driver, and it appears there are a quite a few newer ones, like 7.6.15.5 http://sourceforge.net/project/showfiles.php?group_id=42302&package_id=54835 *** EDIT #2 *** ... My Original 4.2.1 was using a version: 7.6.5-NAPI of the e1000 driver... it seems that from 4.2.1 to 4.3beta we have droped a few revisions to an older e1000 driver... weird... yes? Attached is my 421 Syslog (Line 579 shows 7.6.5 version)...
April 5, 200818 yr SuperW2- Please try 4.3-beta6, which is now available. We upgraded to the latest Intel network driver (7.6.15.4). I didn't notice how down-rev the Intel driver was which is included in the latest linux kernel release & can't explain why that is.
April 5, 200818 yr Author SuperW2- Please try 4.3-beta6, which is now available. We upgraded to the latest Intel network driver (7.6.15.4). I didn't notice how down-rev the Intel driver was which is included in the latest linux kernel release & can't explain why that is. Beta6 Appears to have addressed this situation...I just installed Beta 6 and was able to boot, and have the link stay @ 1000 with no drops... Copying a 4.5 GB folder now to the Cache drive (via the user share) took just over 1 minute @ 60-65 MB/Sec (per Windows)....VERY VERY nice! SW2
Archived
This topic is now archived and is closed to further replies.