NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.
Forum Discussion
how to optimize a temp folder for maximum write/read performance ?
I have a fairly large plex media server/library running on my 528x, upgraded to 16gig ram. I am interested in utilzing an existing, or if necessary creating a specialized temporary folder dedicat...
Unfortunately no I have 8x 8tb raid6/x-raid.
Already having such a large and full volume, tends puts a limit on fast smaller file IO, rapid creation/removal scenarios.
I thought about putting an ssd via usb3, but really even SSD have IO and write cycle limits that would be avoided entirely by giving a specific folder or volume ~10 gig of ram write cache.
maximizing (or minimizing, depending on how you look at it) latency performance is the goal, not sure if usb3 would be an improvement over the baseline raid-6 volume.
If you may remember back in the day, with hardware raid controller boards, some could have several gigs of standard ram dimms for write-back-cache to improve IOPS and latency. Most of these had some kind of battery backup to help ensure that cached data got written in case of a power interuption.
write-back cache to a specific folder/volume I believe is what I am looking for, though not sure if that is something that is feasible.
I thought about using /run or /dev/shm (same thing?), which appear to use up to max of 8 out of 16g as tmpfs+swap.
While this might work perfect for normal streaming transcodes, I fear it would be insufficient to cover long or multiple dvr shows/movies which could cause out of memory problems.
reading more on plex forums and google, it looks likes actually at some point the streaming transcode now *requires* sizeof media_file+100meg free space in order to start a transcode.
So this means for plex transcoding bluray or 4k rips, this could potentially be 50+ gig per user that is streaming simultaniously.
I knew I should have gone with 64 gigs of ram, LOL.
for reference, here are 2 bonnie++ results: 528x 16g ram, 8x 8tb raid-6
main media share
Version 1.97 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP sauron 31840M 93 97 256015 46 140674 41 606 95 411216 37 468.0 39 Latency 159ms 125ms 754ms 16456us 629ms 121ms Version 1.97 ------Sequential Create------ --------Random Create-------- sauron -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 16 1962 14 +++++ +++ 4038 41 4503 32 +++++ +++ 5517 60 Latency 69957us 728us 258ms 36138us 8us 60330us 1.97,1.97,sauron,1,1516794851,31840M,,93,97,256015,46,140674,41,606,95,411216,37,468.0,39,16,,,,,1962,14,+++++,+++,4038,41,4503,32,+++++,+++,5517,60,159ms,125ms,754ms,16456us,629ms,121ms,69957us,728us,258ms,36138us,8us,60330usexternal usb3 SSD - ntfs format
Version 1.97 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP sauron 31840M 414 99 150494 15 25991 4 652 99 230739 17 1438 89 Latency 21255us 33084us 55335us 16644us 18269us 248ms Version 1.97 ------Sequential Create------ --------Random Create-------- sauron -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 16 9591 38 +++++ +++ +++++ +++ 22659 82 +++++ +++ +++++ +++ Latency 3103us 967us 8191us 1476us 18us 6771us 1.97,1.97,sauron,1,1516790413,31840M,,414,99,150494,15,25991,4,652,99,230739,17,1438,89,16,,,,,9591,38,+++++,+++,+++++,+++,22659,82,+++++,+++,+++++,+++,21255us,33084us,55335us,16644us,18269us,248ms,3103us,967us,8191us,1476us,18us,6771usi'll try formatting the ssd with different filesystems and post them later.
I'm wondering if you are working to remove a real bottleneck, or a hypothetical.
Are you finding that Plex can't keep up with the real-time transcoding now? If so, how do you know that it's the disk i/o, and not the CPU?
i
good questions.
I suppose the main problem I am trying to resolve is;
- I can record from my network tuner (hdhomerun quattro) fine.
- I can stream regular bluray/dvd/other media fine.
But when I try to do both, I tend to get screwed up recordings and unstable streaming...
DVR recording is highly write intensive, it is taking a raw OTA tv signal in mpeg2 ts format and recording it to disk continuously while whatever show or movie is being recorded. It is not really CPU bound as it is a just a constant write load, which may also include up to 4 simultanious recordings.
Then there is a post-processing commercial removal after each recording has ended, which is moderately more cpu intensive (scans the recorded file looking for commercial markers and re-writes the new stream).
Then there is the normal media streaming from server to client, which is often either a full audio/video transcode, or audio only or video only transcode, which plex does in like 60 second chunks spread across multiple transcoder threads.
Of which there can be multiple people streaming at the same time.
There is also other background and system processing competing for cpu and io at various times, not to mention the overhead of dual redundancy, checksums, and file creation/deletion time on a large and full btrfs file system.
Now obviously that can be a lot going on, and very well may be I need more cpu, which would be either 628x or newer or competitor models, or a dedicated server/pc for plex.
I can't just pop out and replace the cpu and since I am trying to optimize and work with what I have, which is extra ram, I want to reduce or cushion as much IO from hitting the disks, as I possibly can.
NETGEAR Academy
Boost your skills with the Netgear Academy - Get trained, certified and stay ahead with the latest Netgear technology!
Join Us!
