October 27, 200916 yr Author Since you're planning to swap your parity drive, use the chance to see what your write speeds will look like without the parity disk installed. OK just got home from work and pulled out the parity drive.. I grabbed the same 10 file, 10.92gb package that I used in my previous tests to do this test of writing to the unprotected discs. Speeds are as follows with TeraCopy: WD10EADS # 1 : Total Time = 3 minutes 6 seconds for an average transfer speed of 60MB/s WD10EADS # 2 : Total Time = 3 minutes 3 seconds for an average transfer speed of 61MB/s WD10EADS # 3 : Total Time = 3 minutes 8 seconds for an average transfer speed of 59MB/s WD10EADS # 4 : Total Time = 2 minutes 58 seconds for an average transfer speed of 63MB/s WD10EADS # 5 : Total Time = 3 minutes 8 seconds for an average transfer speed of 59MB/s I then did a normal Vista64 Copy & Paste in Windows Explorer without TeraCopy with the following result: WD10EADS # 5 : Total Time = 2 minutes 57 seconds for an average transfer speed of 63MB/s Out of interest I then copied the 10 files to my HTPC with Win7 RC (Build 7229) and did a TeraCopy to the unraid server from there: WD10EADS # 1 : Total Time = 3 minutes 35 seconds for an average transfer speed of 52MB/s WD10EADS # 3 : Total Time = 3 minutes 24 seconds for an average transfer speed of 55MB/s The copies from the Win7 machine were noticeably slower but it is in a different room, connected to a different switch with the two switches connected via a 15 metre cat6 cable, so there could be networking factors affecting its speed. These transfers weren't "peaky" either. The speeds stayed constant pretty much throughout the whole copying process. Going to drop the Samsung in now and run it through preclear before I assign it as parity.
October 27, 200916 yr @Talos: Thanks for giving me the idea to check out TeraCopy! It didn't bring any improvement to my writing speeds, but I liked it nonetheless. It's sleek, and with very useful features. Purko
October 27, 200916 yr Author @purko : no problemo. I actually found TeraCopy through these forums when looking for a file copy utility that would do on the fly verification on the copied file to make sure it wasnt getting corrupted I now have what is possibly a stupid question The samsung passed through preclear overnight with what appeared to be flying colors. I've just assigned it as parity drive in the Devices tab of the web menu but when I've gone back to the Main tab to start the array its reporting that "The replacement disk must be as big or bigger than the original. The original obviously being the 1.5tb Seagate I removed. This means I can't "Start" the array. Will I have to use the "Restore" button to start the array? Will this remove the data I have on the 5xWD10EADS drives or will it just start the parity sync on the new drive, ignoring that its a smaller drive (but still the same size as the other data drives) and leave all my data alone? Bit loathe to hit Restore without seeking advice after seeing all the warnings around the forums of the evils of the restore button
October 27, 200916 yr Do not press restore... It will do NOTHING to make the disk you just assigned as parity bigger. Your replacement "parity" drive must be showing up as smaller than the original "parity" drive. The real question is... is the size really smaller? I'd try the following command on each of your drives to learn what the Linux operating system sees as the reported size of the disks. It is very possible that the new "parity" drive has an HPA using some of its space. In any case, try the following command on each of your disks and report back with the result. hdparm -N /dev/sdX where sdX = sda, sdb, sdc, sdd, etc, or hda, hdb, hdc if you have PATA devices. Joe L.
October 27, 200916 yr Author Your replacement "parity" drive must be showing up as smaller than the original "parity" drive. The real question is... is the size really smaller? Yes it is. I replaced the original 1.5TB seagate parity drive with a samsung 1TB drive. Below are the results of the command you listed (sdb is the new Samsung 1TB i want to try as parity): /dev/sda: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdb: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdd: max sectors = 1953525168/1953525168, HPA is disabled /dev/sde: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdf: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdg: max sectors = 1953525168/1953525168, HPA is disabled
October 27, 200916 yr Your replacement "parity" drive must be showing up as smaller than the original "parity" drive. The real question is... is the size really smaller? Yes it is. I replaced the original 1.5TB seagate parity drive with a samsung 1TB drive. Below are the results of the command you listed (sdb is the new Samsung 1TB i want to try as parity): /dev/sda: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdb: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdd: max sectors = 1953525168/1953525168, HPA is disabled /dev/sde: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdf: max sectors = 1953525168/1953525168, HPA is disabled /dev/sdg: max sectors = 1953525168/1953525168, HPA is disabled Now I see... Yes, this is one of the few times where you will need to press the button labeled "Restore" It is because the superblock data in config/super.dat thinks the array is really 1.5TB and artificially extended all the smaller drives to that length be simulating the last .5TB of space with zeros when calculating parity. Pressing "restore" will do nothing to your data, but it will rename the existing super.dat to super.old, and then re-create a new config/super.dat file based on the existing assigned and working drives. It will then start on calculating parity on their contents. So... You were smart to read and take note of all the warnings we've posted to not press the button labeled "restore." But this is one of the situations where it is warranted. Joe L.
October 27, 200916 yr Author Thanks for that Joe.. Ill set parity to sync while I'm at work and then do some more testing on write speeds when I get home.
October 28, 200916 yr Author Well after letting the Parity Sync run while I was at work I've just done a few more test transfers with the same 10 file, 10.92gb package I did my earlier tests with.. Speeds are as follows, measured with TeraCopy: WD10EADS # 1 : Total Time = 12 minutes 04 seconds for an average transfer speed of 15MB/s WD10EADS # 2 : Total Time = 10 minutes 32 seconds for an average transfer speed of 18MB/s WD10EADS # 3 : Total Time = 11 minutes 02 seconds for an average transfer speed of 17MB/s WD10EADS # 4 : Total Time = 10 minutes 38 seconds for an average transfer speed of 18MB/s WD10EADS # 5 : Total Time = 12 minutes 24 seconds for an average transfer speed of 15MB/s File transfers are again "peaky", hitting reported speeds as high as 38MB/s and as low as 2.3MB/s but overall average throughput is definitely a lot higher with the Samsung 1TB Eco Power in there compared to the Seagate 1.5TB 7200rpm. So now the question is I guess is there something wrong with the Seagate drive? Its a CC1H drive manufactured in August 2009 which I believe is past the firmware/date the problems were meant to be occuring. Or maybe an incompatability with the ICH10R and 1430SA chipsets? Maybe I should try a 1.5TB Samsung/WD 7200rpm? Edit: 1 thing I just noticed is the Samsung drive doesnt respond to the spindown command. It will spindown itself after 1hour but if I actually press the Spin Down button it wont spin down.
October 28, 200916 yr Different drives have different firmware, and manufacturers tweak their firmware for target performance in different environments. Like designing a camshaft for a car -- you can optimize it for city driving at 35mph, but mileage at highway speeds will suck canal water. Or optimize it for highway speeds and then have horrible performance in city driving. Manufacturers select a design that is a compromise to give decent performance in both situations. There are other differences such as buffers, read-ahead algorithms. Some drives may even buffer a whole track (not cylinder) , which will really supercharge unRAID for writes of large files.
October 29, 200916 yr I have unRAid 4.4.2 I am test throught TeraCopy (nice tool , thanks for tip ) copy to my server some files with summary size 12.43GB size (2x2,5GB, 1x8GB). I am attach process of copy and test to/from server. Green bar is upload, red is download - test file. Mean for upload is 21.53 MB/s, mean for download is 44.04 MB/s. My config of server is bellow under line (1xparity disk, 5xdata disk - all is 7200 RPM). Why upload to server too varies? Upload and download from PC Q6600 (overclock 3,2 GHz), 4GB RAM, XP SP3 Thanks for any ideas.
October 29, 200916 yr Now I am upload and download from notebook U7600 (1,2 GHz), 2GB RAM, XP SP3 I am test throught TeraCopy copy to my server some files with summary size 16.82GB size. I am attach process of copy and test to/from server. Green bar is upload, red is download - test file. Mean for upload is 17 MB/s, mean for download is 32 MB/s. Now upload not very varies.
October 29, 200916 yr This all depends on how fast the client can fill the buffer cache on the server. Then it depends on how fast the buffer cache can be flushed to the array.
October 29, 200916 yr Ropo, what's your point? What's my point? In this thread is question "why ulpoad is volatile". This is behavior by design of unRaid server. Is it so?
October 29, 200916 yr Author what is volatile? I think he's trying to say the throughput isnt constant. Just a different word for varied, peaky, stop-start. And yes I believe that because of the way it works where it's copying the the data and setting the parity state the system will always be volatile/peaky when copying to the array. You don't get this usually when writing to an unprotected disc because there is no need to go off and check then write parity. @Ropo My issue wasn't with the peaky write speeds but the very low write speeds I was getting. This has been rectified with the changing out of the Seagate 1.5TB parity drive. Sold it off to a friend today who is very happy with it in their system as a normal disc under Win64.
October 31, 200916 yr Author Talos, what have You disk for parity drive now? Ive got a 1tb Samsung Eco Power in there.. Model number is HD103SI
Archived
This topic is now archived and is closed to further replies.