I have an older TPLink Wireless N router plus three PCs running the three latest LTS support versions of Ubuntu Desktop.

The oldest OS is installed on my fastest computer that I am using as my NFS server for AV editing with FOSS tools KDENLive and Audacity. I record amateur bands with several independent devices, sync up the audio and video, and edit the recording for release to YouTube on my amateur channel. It's pure hobby at this point but maybe a source of income once I improve.

The NFS is running RAID 6 with encryption and works fine over a wired gigabit Ethernet router. I found that an 8K block size worked fairly well, but honestly it's been so long since I moved from my apartment back to my house and had to go wireless that I can't recall the minimal optimizations I did.

The router has 3 antennas and is centrally located in the upstairs hallway where the signal propagates well throughout the property. The wireless cards are Asus AX with separate dipole-ish antennas. I've oriented all the antennas optimally to the best of my ability and get good coverage.

File transfers over wireless are so slow that I literally can't edit over the network now. The audio software keeps crashing and the video software hangs.

I've resorted to using the server as an archival and playback system only. With a large cache, mplayer can play back 1080P okay, both on my edits and my ripped media, but that's it. File transfers are real time speed.

I've been reading up and managed to figure out with the help of linssid, iperf, and system configuration menu that even though my 5G network has about -50dB to -60dB signal strength, it produces stats in the control panel of about 500Mbps on all the computers when I send packets back and forth between them. However, iperf reports about 100Mbps over 5G and about 90Mbps over 2.4G despite the stronger signal strength.

I'm suspecting some alternative limitation to the speed that I haven't yet debugged.

I tried investigating R/WSIZE but it seems that the MTU is set at 1500 because of the wireless devices and the largest block that is reported is 1024 or one hard disk sector. It's really slow. Apparently some ISPs don't take larger MTUs and I'm concerned about breaking something if I tweak it, if that is even possible with my gear.

I'm looking into running a wired network through a separate router located at the server and feeling bummed about it. The path from the laundry room to the living room is clear through the crawlspace but the path to the upstairs bedroom will have to run outside the building. I'll have to hire someone to do it because I'm not physically capable.

Is there anything at all I can do to improve NFS over wireless so I don't have to resort to such measures? I've looked and looked but didn't find any encouraging news.

Is there an alternative protocol that might work better?

I'm budget constrained so that's why I'm doing everything in FOSS.

Thanks so much for any advice.

NFS is nice, but not terribly performant. Things get worse on WiFi, where large block sizes tend to have variable unhappy consequences even when the hardware supports that (most do not). For a more performant storage connection I would go wired 1000BT full duplex if you have the option. Direct (box to box) is best if your devices have the NIC ports, but through a good switch that is not underpowered is also pretty good. I ran my own wire, but if I could not have I would have asked for a quick estimate form a local electrician.

If you cannot go back to wired for that link, you might try using sshfs. It is a bit slower than NFS for most small things, but does not degrade in the same pattern as traffic changes. I cannot be sure it would provide the performance you require, but it is a cheap, easy option to try.

CIFS using SAMBA tools can be more performant than NFS generally, but I have not used it over WIFI. It is also a bit more involved to implement. I would try sshfs first and see what happens.

What is the speed of your WIFI?
That WIFI connection is critical, and if yours is falling back to an earlier (slower) connection standard it could be dragging you down. Standard involved are 802.11a, 802.11b, d, g, n, ac, and 802.11ax. My current devices are 802.11n and 802.11ac, but I have some old equipment that tops out at 802.11g. Your connection will generally use the highest standard supported by both ends, the distance, and the amount of interference in the environment. Each higher standard is faster, but may have less range. (Generalizations, I know, but as accurate as justified by my inability to measure your situation and environment.) If it is falling back there will be a good reason, and identifying it may help us deal with it.

Wired indeed. I use NFSv4, my home [gigabit] LAN is crisscrossed with NFS mounts, Wife even has her home directory on NFS. There are two switches between my desktop and media server and NFS maxes out my NIC, so CIFS certainly could not perform better here. My NICs are all Intel, BTW, there is no better NIC than Intel. I'm a little suspicious about that CIFS advantage anyway, perhaps it was better than NFSv3?

Editing large remote video files can be a problem. Have you tried using a proxy files?

Been considering this. I might store video stages and work on a network drive, but I would never edit them there. Even on local drives, there is advantage in having the fastest hardware, partitioning, and file system. Using a network drive slows everything down to AT BEST the network speeds. Storage speed is less critical, but working storage speeds need to be something on the order of ALL YOU CAN GET.

Even wired (which should be faster and more dependable) networked drives would be less than optimal.
Can you transfer the files to local storage, edit them there, then offload the result to the NFS drive?

https://www.xmodulo.com/how-to-enabl...-on-linux.html
https://access.redhat.com/documentat...ide/fscachenfs

This is the point of proxy files for video editing. Many video editors (at least the ones I've used) have this feature.

I will take your word. I have probably edited a dozen simple animations in my entire life, and the most recent of those over a decade ago.

Note: My experience with Linux is entirely simplistic. I'm not running a server distribution because it's too labor-intensive to climb the command line learning curve. I'm not running Kerberos and I'm running NFS as root. I don't have home directories or anything mounted on the server. That box is entirely NAS presently, although some day I'd like to try more sophisticated configuration, provided that maintaining it isn't too labor-intensive. My prior experience with Unix is limited to what I've done as an electrical engineer as a user in a networked environment decades ago (Solaris). I once multi-booted many distros with Grub V1 to try them out. Haven't done virtualization but I have done Wine emulation.

I don't have any familiarity with some of the tools mentioned here and barely understand the context of what they are and what they do. I'll take them one at a time...

[QUOTE=wpeckham;6383999]For a more performant storage connection I would go wired 1000BT full duplex if you have the option. Direct (box to box) is best if your devices have the NIC ports, but through a good switch that is not underpowered is also pretty good. I ran my own wire, but if I could not have I would have asked for a quick estimate form a local electrician.[quote]

I have a guy who already did some wiring here, a former cable/satellite installer working his way through retirement. I'll ask him if necessary.

I have never run box-to-box. I looked it up and it seems straightforward. The part I'm not so sure about is how I deal with the presence of both wired and wireless routes between boxes simultaneously. If I add direct connection I'd still want the wireless router available for ISP connection plus I'd want box-to-box NFS to fail over to wireless if/when the wire isn't available.

I figured a separate router or switch at the server was probably easier to configure than box-to-box but maybe not a whole lot. The advantage of direct connection is I can immediately connect the server to the HTPC on the main floor and get running with a wire rigged temporarily to test the theoretically best performance without needing to buy anything or drill any holes. I already have enough cable with a female-female connector to bridge the span with two lengths in series. Might try that first. I think the server has two wired ports so I could possibly bridge the two remote devices without any additional hardware, or if there's only one port in the server I could add another port if necessary.

The only traffic here is me. That's basically one user. Thanks for the suggestion, I'll look up sshfs and see what's involved. Maybe give it a test drive. I think going wired might be less labor-intensive for me and better performance.

I thought these hot wireless routers were supposed to get me near-gigabit speeds.

My prior company used SAMBA to bridge Windows to Linux a couple decades ago. That's all I know about it. The brief review at this link:

https://www.techtarget.com/searchsto...le-System-CIFS

says CIFS is obsolete, buggy, high latency, and labor-intensive.

I have the TP-Link Archer C2300 V.1 wireless N/AC router.

https://www.tp-link.com/us/home-netw.../archer-c2300/

"WIRELESS
Standards Wi-Fi 5
IEEE 802.11ac/n/a 5 GHz
IEEE 802.11n/b/g 2.4 GHz
WiFi Speeds AC2300
5 GHz: 1625 Mbps (802.11ac)
2.4 GHz: 600 Mbps (802.11n)
WiFi Range 3 Bedroom Houses
3× Detachable High-Performance Antennas
Multiple antennas form a signal-boosting array to cover more directions and large areas

Beamforming
Concentrates wireless signal strength towards clients to expand WiFi range
WiFi Capacity Medium
Dual-Band
Allocate devices to different bands for optimal performance

3×3 MU-MIMO
Simultaneously communicates with multiple MU-MIMO clients

Airtime Fairness
Improves network efficiency by limiting excessive occupation
Working Modes
Router Mode
Access Point Mode
HARDWARE
Processor 1.8 GHz 64 Bit Dual-Core CPU
Ethernet Ports 1× Gigabit WAN Port
4× Gigabit LAN Ports

Static Link Aggregation (LAG) available with 2× LAN ports"

The 2.4G radio is configured for N only (top speed) with the wideband setting of 40MHz at high power, airtime fairness.

https://www.linuxquestions.org/quest...1&d=1664989444

The 5G radio is configured for AC only (top speed) with the wideband setting of 80MHz, airtime fairness, multi-user MIMO.

https://www.linuxquestions.org/quest...1&d=1664989561

The firmware stopped updating years ago and this router is obsolete with no new security patches, but still works okay.

(OOPS!! They just released a bug patch for the first time in years and I upgraded to the latest while typing this)

I would consider upgrading it if there were significant speed advantages for the price. Something tells me the router is 'not the problem' or upgrading won't fix anything because wireless AC is already squeezing most of the throughput from the available bandwidth. Mesh probably won't help either. Just guessing here.

There is no mission-critical data but I have a lot of home video recordings and ripped media on the server plus some personal documents and projects I'm fond of.

I tried manually configuring channels for a time, but found that leaving it up to the router worked better with the local traffic. I'm in a cluster of three cul-de-sacs and there are a lot of wireless routers nearby.

Here are the signal strengths at my home theater PC. The top two are my 2.4G network? and the next two are my 5G network?

https://www.linuxquestions.org/quest...1&d=1664987208

Here are the channel allocations for 5G (the spectrum is more congested on the other side of the house with the server in the laundry room and bedroom PC upstairs):

https://www.linuxquestions.org/quest...1&d=1664987527

2.4G channels with a wireless printer upstairs directly above the home theater PC:

https://www.linuxquestions.org/quest...1&d=1664987294

Attached Thumbnails

         

Continued...

I changed my wireless passwords since I exposed them so no worries there lol.

I see SSIDs in linssid that I don't understand, they are very long and begin with w00x00x00x00x00...

The channel and signal strength match my router

https://www.linuxquestions.org/quest...1&d=1664987527

but the MAC addresses don't. I'm wondering what these ghost transmitters are?

The wireless adapters in the PCs are Asus PCE-AX58BT

" Data Rate
802.11a : up to 54 Mbps
802.11b : up to 11 Mbps
802.11g : up to 54 Mbps
WiFi 4 (802.11n) : up to 300 Mbps
WiFi 5 (802.11ac) : up to 1733 Mbps
WiFi 6 (802.11ax) (2.4GHz) : up to 574 Mbps
WiFi 6 (802.11ax) (5GHz) : up to 2402 Mbps
Operating Frequency
2.4 GHz / 5 GHz
OS Support
OS Support : WindowsR 10 64-bit
Linux"

https://www.asus.com/us/Networking-I...es/PCE-AX58BT/

Based on the advertised speeds of the devices, if iperf tells me they are capable of only 100Mbps there's something wrong? Is it signal strength? The bedroom PC doesn't seem to perform any better. They all report about 500Mbps in system control panel on 5GHz after I fire up iperf that only reports 100Mbps between boxes no matter which two are communicating.

The server is in the laundry room downstairs aimed past the rear of the washer/dryer stack and furnace at that router upstairs and on the other side of the main load-bearing wall. The washer/dryer is wifi capable but not connected to wireless. I don't see any signals from it in linssid.

Don't know what a proxy file is. Presumably it's a local mirror of the remote video file?

The most sophisticated video editor I have used is KDENLive that is primitive and underpowered but I'd like to learn something better soon. After I finish editing it and moving it back to the server I have to redirect it at the server copy of the video files because the links are all broken. Fortunately it's a relatively painless and quick adjustment, but it also means I have to move the local video files so it can't find them (with the workflow I've hacked together anyway).

These are 3 hour shows with 3 or 4 camcorders at 1080p and a separate synced and merged audio track, so yes probably large when taken in context of wireless editing.

That's what I have been resorting to. I recently ran out of space on the local file system after I upgraded the OS to my workspace drive, so I had to mount another workspace drive to resume working. I had the drive already installed but hadn't been using it. It's very old (2006) and I'm afraid it will die on me. It used to be one of a 4x1.5TB RAID on an old iteration of the server and the three other drives from that prior server already failed.

Thing is, the NFS is so abominably slow that even when I edit locally, it takes a full day to archive to the server. That's at least four times as long as it took when wired.

Maybe edit the project on the server and link to a local copy of the video files? Maybe I can set up an NFS server on the local machine to serve the temporary copy of the video files back to the server through a symbolic link? Then all I have to do is toggle one link in the project directory on the server to toggle between local and remote raw video files, while keeping the full archive of the project files on the server so all the paths remain consistent. Then I don't have to edit the project to resume working on it, I just have to copy the video locally and redirect the paths.

Um...is this doable/practical? It seems like it could really get me out of a jam. Is that how I use 'proxy files'? with software that doesn't support such a thing? Far as I know, all the paths are absolute. I'll look at this closer in KDENLive.

Final edited video plays OK from the server with the network as is so I'm covered there.

KDENLive proxy clips are low resolution copies.

https://kdenlive.org/en/project/kdenlive-proxy-clips/

I'd rather not generate more data. I'm already space limited.

I've come up with a plan for KDENLive. I'll duplicate the remote directory heirarchy of the NFS server on a local drive in the root of the local machine and keep a local copy of the raw video there. The local volume will be a manually generated and sparsely populated mirror of the NFS volume.

I'm already binding the local mount point of the NFS volume from /mnt/archive to /archive in fstab so that the paths on the file server and the local machine are consistent.

I'll create a symbolic link to /archive named /raw on both the server and on the local machine and add all raw files to the project through the path at /raw. Then the project integrity is preserved in local or remote contexts without needing to make any changes inside the project and I've got a symbolic link on the local machine to redirect to the local files when needed for editing.

To verify project integrity, I'll have two options.

I can open the project on the server and check that the editor finds everything. If so, all is well.

Or, I can switch the /raw link on the local machine from the local volume /data to the remote volume /archive and check that the editor finds everything. If so, all is well.

I know, it's kludgy, but it should work okay and it's simple. I think this is where I'm headed. I don't want to run new wires or mess around with software changes for potentially minimal performance gains (at least not until I'm trying to play back 4K video from the server, something I haven't yet attempted).

Audacity uses relative paths for linking its project data to its project file. I can edit a project locally and then copy it to the server without losing any paths. I can also edit it on the server, although it's very slow. For minor post-processing tweaks I won't have to copy it locally.

It's not as good as being able to efficiently edit remotely, but it's okay.

Something else I'm thinking about is VNC Server. I've used that before to edit remotely across continents. It should work okay over wireless LAN, but the display isn't 1-for-1 so there can be glitches with the GUI of applications that run remotely. Since the NFS server is also the fastest machine in the house, it would be preferable to edit on the server, especially when working in the bedroom that is a much slower machine. I haven't set this up to try it yet because, well, it's even more of a kludge using a remote display than copying files locally. I might give it a try anyway.

Thanks for the suggestions. I'll mark this one solved.

And/or a lower quality version of the file. See for example
https://jliljebl.github.io/flowblade/webhelp/proxy.html

Evo2.

They're temporary files that are only needed during the editing process
(assuming kdenlive implements this in a non-braindead way).

Evo2.