March 4, 201214 yr It's just a script to test your direct disk throughput. You can run it on the free version of unRAID. after you download it, do a chmod u+x ./writeread10gb Then run it with ./writeread10gb /mnt/disk?/test.dd where ? is the disk you want to test 1-20. It will print some output. At that point you know your maximum disk throughput for that device and parity.
March 5, 201214 yr Author I know what won't work: I made the 3 TB volume my parity disk and the 500 GB volume a data disk, parity check goes at just below 10 MB/sec. Should I have precleared the the drives/volumes or does that make no sense? When I unassign the parity disk and assign the 500 GB volume to the disk1 slot and run the readwrite10gb scriptt on it what speeds can be expected then?
March 5, 201214 yr You don't want the 500g drive part of the array as its on the same disks as your parity. Why your seeing really slow parity speeds. Sent from my SGH-I727R using Tapatalk
March 5, 201214 yr You don't want the 500g drive part of the array as its on the same disks as your parity. Why your seeing really slow parity speeds. The 500GB RAID1 volume should not be part of the array. It's already protected via RAID1. Plus it will compete for access to the same two drives. Use it as an APPS drive or a cache drive if you upgrade.
March 5, 201214 yr Author I know, I know. It was just a try-out. I now added a Samsung 1 TB drive. Parity sync is in progress now, but is going slow, around 20 MB/sec. Did I do something wrong configuring the ARC1200, Raid set and Volumes? Are there settings that influence speed? Stripe size perhaps? SATA300 or SATA300 and NCQ?
March 5, 201214 yr Author To be honest, I don't know if WB-cache is enabled. I haven't change any settings, is it enabled by default? No, I don't have a UPS, is it a neccessity?
March 5, 201214 yr If you enable the WB cache it's a necessity. Frankly, I think a UPS on a file server is a necessity no matter what. The Write Back cache is not normally enabled, you have to turn it on. In any case 20MB/s is too slow. Enable it and see how you perform.
March 5, 201214 yr Author What is the risk if you enable wb cache without ups? How big is the speed improvement? I understand your comment about a UPS. Which UPS is recommended?
March 5, 201214 yr Author If you enable the WB cache it's a necessity. Frankly, I think a UPS on a file server is a necessity no matter what. The Write Back cache is not normally enabled, you have to turn it on. In any case 20MB/s is too slow. Enable it and see how you perform. About the cache: I only have 2 options: Write-trough or Write Back. There is no option to disable cache. The setting was Write Back and still is. I changed something else though: When I selected to create a Volume I was asked to enable >2 TB support I had 3 options: No, 64 bit LBA, 4K Block. Yesterday I chose 4 K Block, this time 64 bit LBA. Parity is now running (just started, at 2%) at 71.88 MB/sec. Lesson learned: When enabling > 2 TB support choose 64 bit LBA!
March 6, 201214 yr Thanks for your input on this! For a UPS I would recommend an APC since apcupsd can be installed from unmenu.
March 6, 201214 yr Author Will any APC do? Back-UPS or SMART-UPS? What power/wattage would I need for a system comparable to yours. 900VA? How much run-time would I need for a clean powerdown of the server?
March 6, 201214 yr My UPS's are old SUNET1400's. I don't have a recommendation for a modern one. I think you should look in the other threads or post a new one.
March 6, 201214 yr Author I guess I need to run in one. Now about speed: Was the Parity check running at about 72 MB/s ok? I also did the writeread10gb test. Writing was about 35 MB/s. The readtest at the end 98 MB/s. Does that seem ok? I don't have a cache drive yet, because I'm on unraid free. Can I make any changes to improve speed?
March 6, 201214 yr Those numbers are good. You can improve it slightly depending on how much ram you have. In disk settings I doubled them. Tunable (md_num_stripes): 3840 user-set Tunable (md_write_limit): 2304 user-set Tunable (md_sync_window): 864 user-set I also adjusted other settings in the kernel to allow top speed bursts for small file usage on a file server basis. Since my source is on NFS, it was important to me, YMMV. My adjustments are sysctl -w vm.dirty_ratio=40 sysctl -w vm.dirty_writeback_centisecs=50 sysctl -w vm.dirty_background_ratio=20 sysctl -w vm.min_free_kbytes=8192 I would suggest googling each setting so you understand what is changing. If you want to discuss these settings in more detail, start a new thread. let's keep this one specific tot he ARC-1200 Setup. We can always add other links into this thread for someone to reference other conversations.
March 6, 201214 yr Author Are those settings in unraid itself or ARC1200 specific? I only have 4 GB RAM.
March 6, 201214 yr Are those settings in unraid itself or ARC1200 specific? I only have 4 GB RAM. unRAID, 4GB is enough.
March 6, 201214 yr Author Right now a Parity check is running at close to 190 MB/s at 65%. I think yesterday was no Parity check but a Parity sync at about 72 MB/s, right after the parity disk and a data disk were added. What is the difference between Parity sync and Parity check?
March 7, 201214 yr Parity check reads all disks and compares parity. This is usually slightly slower because the cache does not aid as much. Parity sync creates parity, the wb cache comes into play and aids in speed because the buffer is constantly revolving and rotational delays are minimized. 190MB/s @ 65% is because only the parity drive is being used. 65% of 3TB when you only have 1 1tb drive means you've exceeded the need to read the 1TB drive.
March 7, 201214 yr Author Logic explanation. Could have invented that one myself. At 65 % I didn't see any reads from disk1, in fact disk 1 was spun down. Now I have to find where I can set spin down on the ARC1200.
March 7, 201214 yr Logic explanation. Could have invented that one myself. At 65 % I didn't see any reads from disk1, in fact disk 1 was spun down. Now I have to find where I can set spin down on the ARC1200. It's in the bios. I think the max was 1 hour. I wrote a tool to keep the drives spinning all day, then let it spin down at night.
March 7, 201214 yr Author Weebo, Just added the cache drive, which is the RAID1 volume on the ARC1200, to the server. writeread10gb gives these results: root@Hal:/boot# writeread10gb /mnt/cache/test.dd writing 10240000000 bytes to: /mnt/cache/test.dd 351833+0 records in 351833+0 records out 360276992 bytes (360 MB) copied, 5.02428 s, 71.7 MB/s 635441+0 records in 635441+0 records out 650691584 bytes (651 MB) copied, 10.0464 s, 64.8 MB/s 971188+0 records in 971188+0 records out 994496512 bytes (994 MB) copied, 15.4632 s, 64.3 MB/s 1290909+0 records in 1290909+0 records out 1321890816 bytes (1.3 GB) copied, 21.4749 s, 61.6 MB/s 1524148+0 records in 1524148+0 records out 1560727552 bytes (1.6 GB) copied, 25.1399 s, 62.1 MB/s 1813101+0 records in 1813101+0 records out 1856615424 bytes (1.9 GB) copied, 31.5114 s, 58.9 MB/s 1991764+0 records in 1991764+0 records out 2039566336 bytes (2.0 GB) copied, 35.1552 s, 58.0 MB/s 2339341+0 records in 2339341+0 records out 2395485184 bytes (2.4 GB) copied, 41.1692 s, 58.2 MB/s 2619089+0 records in 2619089+0 records out 2681947136 bytes (2.7 GB) copied, 45.2011 s, 59.3 MB/s 2861533+0 records in 2861533+0 records out 2930209792 bytes (2.9 GB) copied, 50.7844 s, 57.7 MB/s 3225817+0 records in 3225817+0 records out 3303236608 bytes (3.3 GB) copied, 55.2491 s, 59.8 MB/s 3517309+0 records in 3517309+0 records out 3601724416 bytes (3.6 GB) copied, 61.1915 s, 58.9 MB/s 3767190+0 records in 3767190+0 records out 3857602560 bytes (3.9 GB) copied, 65.2878 s, 59.1 MB/s 4039501+0 records in 4039501+0 records out 4136449024 bytes (4.1 GB) copied, 70.8316 s, 58.4 MB/s 4302621+0 records in 4302621+0 records out 4405883904 bytes (4.4 GB) copied, 75.437 s, 58.4 MB/s 4679964+0 records in 4679964+0 records out 4792283136 bytes (4.8 GB) copied, 80.3478 s, 59.6 MB/s 4824813+0 records in 4824813+0 records out 4940608512 bytes (4.9 GB) copied, 85.4197 s, 57.8 MB/s 5217469+0 records in 5217469+0 records out 5342688256 bytes (5.3 GB) copied, 91.5805 s, 58.3 MB/s 5480589+0 records in 5480589+0 records out 5612123136 bytes (5.6 GB) copied, 95.8354 s, 58.6 MB/s 5684959+0 records in 5684959+0 records out 5821398016 bytes (5.8 GB) copied, 100.848 s, 57.7 MB/s 6002781+0 records in 6002781+0 records out 6146847744 bytes (6.1 GB) copied, 106.213 s, 57.9 MB/s 6265901+0 records in 6265901+0 records out 6416282624 bytes (6.4 GB) copied, 110.576 s, 58.0 MB/s 6570222+0 records in 6570222+0 records out 6727907328 bytes (6.7 GB) copied, 115.492 s, 58.3 MB/s 6788093+0 records in 6788093+0 records out 6951007232 bytes (7.0 GB) copied, 120.725 s, 57.6 MB/s 7051213+0 records in 7051213+0 records out 7220442112 bytes (7.2 GB) copied, 125.997 s, 57.3 MB/s 7420418+0 records in 7420418+0 records out 7598508032 bytes (7.6 GB) copied, 130.552 s, 58.2 MB/s 7616161+0 records in 7616161+0 records out 7798948864 bytes (7.8 GB) copied, 135.577 s, 57.5 MB/s 7966061+0 records in 7966061+0 records out 8157246464 bytes (8.2 GB) copied, 141.282 s, 57.7 MB/s 8256729+0 records in 8256729+0 records out 8454890496 bytes (8.5 GB) copied, 145.622 s, 58.1 MB/s 8492301+0 records in 8492301+0 records out 8696116224 bytes (8.7 GB) copied, 151.04 s, 57.6 MB/s 8763660+0 records in 8763660+0 records out 8973987840 bytes (9.0 GB) copied, 155.671 s, 57.6 MB/s 9091192+0 records in 9091192+0 records out 9309380608 bytes (9.3 GB) copied, 160.693 s, 57.9 MB/s 9395409+0 records in 9395409+0 records out 9620898816 bytes (9.6 GB) copied, 165.718 s, 58.1 MB/s 9600999+0 records in 9600999+0 records out 9831422976 bytes (9.8 GB) copied, 170.741 s, 57.6 MB/s 9933389+0 records in 9933389+0 records out 10171790336 bytes (10 GB) copied, 176.963 s, 57.5 MB/s 10000000+0 records in 10000000+0 records out 10240000000 bytes (10 GB) copied, 177.367 s, 57.7 MB/s write complete, syncing reading from: /mnt/cache/test.dd 10000000+0 records in 10000000+0 records out 10240000000 bytes (10 GB) copied, 142.862 s, 71.7 MB/s removing: /mnt/cache/test.dd removed `/mnt/cache/test.dd' Your opinion?
March 7, 201214 yr Author Ok, thanks. One thing I noticed: on the mymain-page both volumes show as HPA.
Archived
This topic is now archived and is closed to further replies.