May 15, 201313 yr Thanks for the explanation, do you have any links to the patch cable you described, I'd like to buy some and proceed with testing. Thanks! I've had the Cat-5 and Cat-5e cables for years. I bought the Cat-6 cable here: http://www.monoprice.com/products/product.asp?c_id=102&cp_id=10232&cs_id=1023215&p_id=2119&seq=1&format=2
May 15, 201313 yr Author Thanks for the explanation, do you have any links to the patch cable you described, I'd like to buy some and proceed with testing. Thanks! I've had the Cat-5 and Cat-5e cables for years. I bought the Cat-6 cable here: http://www.monoprice.com/products/product.asp?c_id=102&cp_id=10232&cs_id=1023215&p_id=2119&seq=1&format=2 Just to clarify this is not a patch that I will connect to a CAT5 cable to improve its performance, but to replace CAT5 cables...
May 15, 201313 yr Thanks for the explanation, do you have any links to the patch cable you described, I'd like to buy some and proceed with testing. Thanks! I've had the Cat-5 and Cat-5e cables for years. I bought the Cat-6 cable here: http://www.monoprice.com/products/product.asp?c_id=102&cp_id=10232&cs_id=1023215&p_id=2119&seq=1&format=2 Just to clarify this is not a patch that I will connect to a CAT5 cable to improve its performance, but to replace CAT5 cables... There is nothing you can connect to a cable to improve its performance. A patch cable is a cable with terminals on each end. Yes, it is a replacement.
June 2, 201313 yr Author Now testing cat5e which I found lying around, I'm getting 22MB writes per second (copied across more than 10 gigs). I've ordered some Cat6 for further testing, I expect write speed to improve further. Thank you all for your suggestions Sent from my GT-I9300 using Tapatalk 2
June 6, 201313 yr Author Now that I have CAT6 my maximum write speed to the same unraid drive (a 2GB WD) is 15MB/second. The drive is only 70% full. Mmmh. This speed is reported in the transfer window using windows explorer. Is there a better way to test writing speed when moving data across the network? The tests are all run between PC with gigabyte card connected with Cat 6 to a gigabit router, connected to another gigabit router via Cat 6 connected to the Unraid server with CAT6 cable. Is this speed affected by the fact that parity (I guess) is also being written on the fly? How do I turn that off?
June 6, 201313 yr Is this speed affected by the fact that parity (I guess) is also being written on the fly? How do I turn that off?You don't. Parity constantly kept up to date is a feature of unraid. To bypass that, you need to be writing to the cache drive, which was introduced specifically to create a non-protected area for fast writes.
June 6, 201313 yr Author What transfer speed should I expect with my hardware setup on a gig network? Sent from my GT-I9300 using Tapatalk 2
June 6, 201313 yr What transfer speed should I expect with my hardware setup on a gig network?I think you are probably seeing close to the fastest speeds that can be expected from the hardware you list. Especially if you have more than 1 drive connected to each of your PCI cards.
June 6, 201313 yr What transfer speed should I expect with my hardware setup on a gig network? Sent from my GT-I9300 using Tapatalk 2 If writing to the protected array, about 20 to 30 MB/s sustained is usual IF the system sending the data can sustain that rate. Some people will report faster speeds on smaller files, but in most cases they are writing to the disk buffer cache. If the source PC only has a 100Mb/s LAN port, then best you'll see is 10 to 12 MB/s. If writing to the cache drive, you can expect faster speeds, but again limited by the LAN and the source PC. Joe L.
June 7, 201313 yr Author Thanks Joe, The PC that is sending the data has a Gigabit port, and the Lan is made up by Giga switches and CAT6 network. I tested the writing speeds to each of the drives this morning, writing a 3.75 GB mkv file to each disk, and recording the write speed. The file is on one of the hard disks of one PC and is getting written on the each drive of my unraid setup. Speeds recorded below (rounded to the lowest number when fractions above a full MB) Disk Mounted Device Model Max Write MB/second Size Used %Used Free Parity ST4000DM000 n/a 4.00T n/a n/a n/a /dev/md1 /mnt/disk1 /dev/sdn SAMSUNG_HD204UI 27 2.00T 1.88T 94% 120.98G /dev/md2 /mnt/disk2 /dev/sdp Seagate ST3500630AS 21 500.09G 470.99G 95% 29.11G /dev/md3 /mnt/disk3 /dev/sdo Seagate ST3500630AS 21 500.09G 471.59G 95% 28.50G /dev/md4 /mnt/disk4 /dev/sdg Seagate ST3500641AS 17 500.09G 468.83G 94% 31.26G /dev/md5 /mnt/disk5 /dev/sdh Seagate ST3500641AS 17 500.09G 471.39G 95% 28.71G /dev/md6 /mnt/disk6 /dev/sdl Seagate ST3500641AS 17 500.09G 461.13G 93% 38.96G /dev/md7 /mnt/disk7 /dev/sdi Seagate ST3500641AS 17 500.09G 471.09G 95% 29.00G /dev/md8 /mnt/disk8 /dev/sdj Seagate ST3500641AS 12 500.09G 467.73G 94% 32.36G /dev/md9 /mnt/disk9 /dev/sdk Seagate ST3500630AS 21 500.09G 465.13G 94% 34.96G /dev/md10 /mnt/disk10 /dev/sdm Seagate ST3500630AS 21 500.09G 459.20G 92% 40.89G /dev/md11 /mnt/disk11 /dev/sdf Seagate ST3320820AS 18 320.06G 300.29G 94% 19.77G /dev/md12 /mnt/disk12 /dev/sde WDC_WD20EARS 11 2.00T 1.84T 93% 155.44G /dev/md13 /mnt/disk13 /dev/sdd WDC_WD20EADS 21 2.00T 1.52T 77% 476.48G /dev/md14 /mnt/disk14 /dev/sdc WDC_WD20EARS 7 2.00T 143.86G 8% 1.86T So different writing speeds for different drives, from 7MB/sec for my emptiest drive to 27MB/sec for the I wonder if this has to do with - are the disks connected to the motherboard - are they connected to the Promise card - are they connected to the same Promise card where parity is connected (if it is connected to a card) I have no idea which drive is connected with which SATA port, I will map things through tonight.
June 7, 201313 yr Replacing PCI with PCIe will dramatically increase access rates. All of the PCI slots combined can only support a single drive with acceptable speed. PCIe slots are independent. PCIex1 - 2 drives PCIex2 - 4 drives PCIex4 - 8 drives PCIex16 - 32 drives Double these number for PCIe 2.0. Subtract 1 drive to get maximum speed from fast modern drives.
June 7, 201313 yr Author Im using 2 simple PCI for the 2 Promise Ultra 100 TX2. The board also has 1 x PCI Express x16 3 x PCI Express x1 Which card can I buy keeping in mind I'm using the 8 x SATA II 3Gb/s connections on the motherboard already. Basically I need another 12 Sata connections to support the maximum 20 drives the case can contain.
June 7, 201313 yr Author http://lime-technology.com/guidelines-for-system-builders/ According to this page "For a less expensive solution, you can use PCI disk controllers. Your parity sync/check speed will not be as fast, but all other normal operations (read/write of media data) will be just as fast. Since parity sync/check can be done during times when your server is not being used, parity sync/check speed is less important."
June 8, 201313 yr http://lime-technology.com/guidelines-for-system-builders/ According to this page "For a less expensive solution, you can use PCI disk controllers. Your parity sync/check speed will not be as fast, but all other normal operations (read/write of media data) will be just as fast. Since parity sync/check can be done during times when your server is not being used, parity sync/check speed is less important." That page is out of date.
June 8, 201313 yr To see what your basic buffer cache/drive throughput is, log in to your unRaid box via telnet and run hparm -tT /dev/sd* | tee ./hdspeed.txt and copy/paste or attach the .txt file in a reply. This does not include filesystem overhead or network overhead/issues. I attached mine with each drive identified as an example of what you will get from it. From http://linux.die.net/man/8/hdparm: -T Perform timings of cache reads for benchmark and comparison purposes. For meaningful results, this operation should be repeated 2-3 times on an otherwise inactive system (no other active processes) with at least a couple of megabytes of free memory. This displays the speed of reading directly from the Linux buffer cache without disk access. This measurement is essentially an indication of the throughput of the processor, cache, and memory of the system under test. -t Perform timings of device reads for benchmark and comparison purposes. For meaningful results, this operation should be repeated 2-3 times on an otherwise inactive system (no other active processes) with at least a couple of megabytes of free memory. This displays the speed of reading through the buffer cache to the disk without any prior caching of data. This measurement is an indication of how fast the drive can sustain sequential data reads under Linux, without any filesystem overhead. To ensure accurate measurements, the buffer cache is flushed during the processing of -t using the BLKFLSBUF ioctl. For more info on speeds you can expect with fairly modern hardware, I did a post a while back comparing write speeds of SMB vs NFS using disk-disk, RAM-disk, RAM-RAM, etc. as explained in the linked post. In a nutshell, you can have fast disks and your throughput will suffer from a slow network and vice versa. Unraid 5.0-rc10 - Asus M5A78L-MLX Plus (RT8111E, 6x SATA2) - AMD Athlon II X3 450 Rana 3.2GHz - 8GB DDR3 - Antec NEO ECO 620W - Antec Three Hundred Case - 1x Rosewill RC-211 (2x SATA2) - 1x Syba SY-PEX40039 (2x SATA3) - Parity: 3T Seagate ST3000DM001 - DATA: 1x 3T Seagate ST3000DM001 - 2x 1T Seagate ST1000DM003 - 2x WD Black 750G - Cache: 250 WD RE3 250G w/8GB swap partition hdspeed.txt
June 8, 201313 yr Author Thanks unevent. I will perform the tests tonight. So far I switched parity to the motherboard sata ports, and changed sata port for disk14 (on the same promise card), reseated cable etc and now I have 11MB/sec speed on that drive at least.
June 8, 201313 yr http://lime-technology.com/guidelines-for-system-builders/ According to this page "For a less expensive solution, you can use PCI disk controllers. Your parity sync/check speed will not be as fast, but all other normal operations (read/write of media data) will be just as fast. Since parity sync/check can be done during times when your server is not being used, parity sync/check speed is less important." That page is out of date. Very out of date. With modern drives, it should read "... Since parity sync/check can be done during the weeks of the month when your server is not being used, parity sync/check speed is less important." The server will be mostly unusable for at least a week a month.
June 8, 201313 yr Author http://www.scan.co.uk/products/highpoint-rocket-2720sgl-internal-8x-sata-sas-6gb-s-hbas-pci-e Would this card work in my PCIe X1 slot? Would I still get a sensible speed improvement? Sent from my GT-I9300 using Tapatalk 2
June 8, 201313 yr http://www.scan.co.uk/products/highpoint-rocket-2720sgl-internal-8x-sata-sas-6gb-s-hbas-pci-e Would this card work in my PCIe X1 slot? Would I still get a sensible speed improvement? Sent from my GT-I9300 using Tapatalk 2 1. No one seems to have really tested that card with unraid to confirm it works and 2. trying to use that card in a 1x slot would result in terribly slow performance. It looks like your motherboard is only PCIe 1.0 so you'd be sharing a single 250MB/s PCIe 1.0 lane with 8 disks. That would get each disk a little over 31MB/s. Modern rotating disk hard drives are able to reach upward of 180MB/s read/write, to put it into perspective.
June 9, 201313 yr http://www.scan.co.uk/products/highpoint-rocket-2720sgl-internal-8x-sata-sas-6gb-s-hbas-pci-e Would this card work in my PCIe X1 slot? Would I still get a sensible speed improvement? Sent from my GT-I9300 using Tapatalk 2 As mrow already stated, putting that on x1 would not be good, though you can try it in the x16 slot and see if it works. Some older boards had the x16 for video only so might want to verify the return policy with the company you are buying from just in case. Also, do a forum search for the part number to see if anyone else has used it with unRaid. The card does show Linux support, so as long as it can be configured as an HBA with no RAID BIOS then it should worktm. Might help to use one of the later unRaid releases (rc10-13) for the newer Linux kernel for driver support. While you're in there, increasing the amount of RAM to say 2-4GB would be a benefit as well.
June 9, 201313 yr Author http://www.scan.co.uk/products/supermicro-8-port-sas-sata-card-(aoc-sas2lp-mv8) How about this Supermicro one? I agree with trying the x16 port, also because the x1 seems to be too small. What does "JBOD mode only, no Raid support" mean? I will buy more Ram as well, thanks for the tip. Btw i'm on unRAID OS version 5.0-rc12a This is the layout of my mboard
June 10, 201313 yr Author Just a question that has been bugging me. I have 8 sata ports on the mboard. I am using two for - parity - disk14 Why would a data transfer to disk14 be influenced by how "slow" the PCI connectivity to all the other disks connected to my 2 Promis Ultra PCI expansion cards? PS: I have just ordered the AOC-SAS2LP-MV8, which I will test on the PCIe x16 slot. This will take care of the 7 drives not currently directly plugged in the SATA connections on the motherboard.
June 10, 201313 yr Author Also, I've been reading about different PCIe standards here http://en.m.wikipedia.org/wiki/PCI_Express As my board offers a PCIe x16 v1.0, and the card is a PCIe x8 v2.0, if, as I hope, it is backward compatible, it should give me 250MB transfer per hard disk. Am I right? Sent from my GT-I9300 using Tapatalk 2
Archived
This topic is now archived and is closed to further replies.