January 29, 20251 yr Community Expert I'm running a 10G network 6 computers in a large rack. NICs are all 10G, Switch is 10G, Cat8 cable. I frequently copy 20-50GB files from one PC to the NAS and they start out fast 700-800MB/sec for about 1 minute or so then thruput suddenly drops to 190MB/sec and then drop again to 60MB/sec and sometimes even 20MB/sec. If I perform the same file copy PC to PC, I will sustain 800MB/sec without much of a change in thruput. When I copy a file from PC to NAS, the performance degradation seems directly proportional to file size. If a file is under 2GB or so, it's very fast, no thruput drop. My NAS (UnRaid v7.0.0 but issue was the same with V6.x.x) has a 3TB Cache Pool (very fast M.2s + some fast SSDs total 6TB). The shares are setup Cache --> Array so my assumption is writing to Cache which should be much faster than 1200MB/sec (10G limit). Plenty of cache space available when I do the copies (2+ TB) so I'm puzzled at the thruput performance degradation based on file size? Rob. Edited January 29, 20251 yr by RobAinscough
January 29, 20251 yr Community Expert 2 hours ago, RobAinscough said: I frequently copy 20-50GB files from one PC to the NAS and they start out fast 700-800MB/sec for about 1 minute or so then thruput suddenly drops to 190MB/sec and then drop again to 60MB/sec and sometimes even 20MB/sec. That suggests the devices are not able to keep up with the sustained writes.
January 29, 20251 yr Author Community Expert The specifications of the drives (both Cache and Array) are well above 60MB/s and 20MB/s and "sustained writes" The Cache Pool -------------- 1X WDS200T3X0E (WD Black SN770) is 4850MB/s for first 325GB (SLC cache full) then drops to 500MB/s for rest of duration (sustained). 2X Lexar SSD NS100 2TB is 467 MB/s sustained The Array Drive --------------- 2X Seagate ST10000NM0016 10TB is 254 MB/s sustained 3X Western Digital WDC_WD101EFBX 10TB is 215 MB/s sustained 5X Seagate ST20000NM004E 20TB is 285 MB/s sustained The above are "sustained" specifications (not sequential). So something doesn't add up, even my worse case device is 215 MB/s.
January 29, 20251 yr Community Expert Is turbo write enabled? With normal write mode, max write seed will be around 60MB/s, due to parity updating.
January 30, 20251 yr Author Community Expert Not sure, I executed: mdcmd set md_write_method 1 from terminal ... but other than the "Spin Up" button how does one enable Turbo Write? What should this be set to? Reconstruct write? Edited January 30, 20251 yr by RobAinscough
January 31, 20251 yr Author Community Expert Did some more testing ... Source files is a 31GB video file mp4 Local Drive (on different PC Win11): M.2 Samsung 990 Pro 2TB For my share called "Assets" which is Cache --> Array For my share called "Media" (plex uses) is just Array Copy 31GB file from Local to Assets 48.53 sec ... Starts at 1.2GB/s then at the 20 second mark drops down to 234 MB/s by end of xfer. Copy 31GB file from Local to Media 154.16 sec ... Starts at 1.2GB/s then at the 19 second mark drops down to about 64 MB/s. The "Assets" Cache --> Array share is faster (as expected). However, I do NOT understand why both tests have significant drop-offs around the 20 second mark? My next question, does Plex have any issues with "Cache --> Array"? My "Media" share is used by Plex and I heard a "rumor" that Plex doesn't like Cache --> Array on UnRaid? Rob. Edited January 31, 20251 yr by RobAinscough
January 31, 20251 yr Community Expert Solution 4 minutes ago, RobAinscough said: I do NOT understand why both tests have significant drop-offs around the 20 second mark? RAM buffer filled up.
January 31, 20251 yr Author Community Expert RAM Buffer? On UnRaid server? If so, I have 128GB RAM with 123GB RAM free. EDIT: also tested the same local file 31GB xfer to another local M.2 drive and xfer sustained average was 3.8GB/s so there is no sustained read issue from my local. Edited January 31, 20251 yr by RobAinscough
January 31, 20251 yr Author Community Expert Ah, found it ... default is 20% RAM Cache so bumped it to 40% and xfer was much faster. Yes, I'm on a UPS so should be safe. sysctl vm.dirty_ratio=40 I assume I will need to set this on every reboot? Is there away I can make this change permanent?
January 31, 20251 yr Community Expert 55 minutes ago, RobAinscough said: Is there away I can make this change permanent? Tips and Tweaks plugin
January 31, 20251 yr Author Community Expert 1 minute ago, trurl said: Tips and Tweaks plugin Is the plugin safe? Last file manager type plugin I used and it kill my Plex app because it dropped Cache drives and had to uninstall and re-install and reset my cache pool … so I’m a little gun-shy when it comes to plugins.
January 31, 20251 yr Community Expert The tweak you are looking for is in that plugin. You don't have to change anything else. 5 minutes ago, RobAinscough said: Last file manager type plugin I used and it kill my Plex app because it dropped Cache drives Unclear what exactly you are talking about or what happened, but your cache problem must have been coincidental.
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.