@boboki said:
Currently PRT pretty much works like this (3rd party distributive). You select a video to play, then the PMS dings all your xcode boxes (ssh session and uses that ssh to xfer too) and finds out which one is the least busy, and starts the xcode on that box. It mostly works right now, and if Plex wants to use that setup for their own, I am betting the guy that made it would be happy to oblige.
I would be happy with that kind of solution if PRT could wake up transcoding boxes over the network and then make the player wait a little bit so the box could start serving. That would allow to spin up and down the transcoding boxes almost like virtual machines.
@boboki said:
Currently PRT pretty much works like this (3rd party distributive). You select a video to play, then the PMS dings all your xcode boxes (ssh session and uses that ssh to xfer too) and finds out which one is the least busy, and starts the xcode on that box. It mostly works right now, and if Plex wants to use that setup for their own, I am betting the guy that made it would be happy to oblige.
I would be happy with that kind of solution if PRT could wake up transcoding boxes over the network and then make the player wait a little bit so the box could start serving. That would allow to spin up and down the transcoding boxes almost like virtual machines.
Well, currently it just issues a SSH⦠but I do not see at all why you could not send a wakeonlan to the boxes before the SSH goes out. Client side it would be a long ole wait before the buffering/movie starts to play⦠but if you were good with that I do not see why that could not be done. The guy that does PRT wnielson has a gitter chat, maybe you could convince him to code that in for like a special release or something⦠until Plex finally does⦠something.
Iām still waiting for this feature! Iād love to be able to shut down the main transcoding server and then just send a WOL-package whenever I need the extra juiceā¦
I think the solution might be to just allow people to purchase more and more h265 HW encoders⦠they are dirt cheap⦠https://www.kickstarter.com/projects/udoo/udoo-x86-the-most-powerful-maker-board-ever
That is a pretty beefy box coming out and it has h265 HW built in, and is sub $100. Would be great to be able to chain a bunch of something like that together to allow transcodes. Just need plex to support it finally.
From the standpoint of our needs that board doesnāt really add much you canāt already get in low end NUCs at present. Even the high end box is only using the Intel 405 chipset on a Pentium processor so itās not going to help you with X.265 encoding. Decoding yes.
QuickSync wonāt be nearly as good as any of the recent chipsets either.
Iād still think a ādedicatedā background client running on a ShieldTV would be a better plug and play client that could also be used in the house connected to a TV for output (2 for 1 use). Not to mention Plex server already works on this platform.
But yea in the near future we will be seeing more and more devices and NUCs that could do real-time x.265 encoding/transcoding.
Basically MAAS will setup a rackfull of servers as transcoding āenginesā then switch them off till needed and wake them using wake on lan, juju is hard to explain watch the video!
I had really hoped that this would be implemented by now.
Here is my use case
I have a 4 core I7 (some mobile chipset) in a small form factor, low power pc that is powered on 24/7. This is sufficient for most things but now that my library is filling up with 4k video, itās not enough CPU anymore.
I have a big gaming PC with a 12 core I7 and an Nvidia 1080i that is very powerful but cannot be left switched on 24/7 and I also need to use itās power for gaming when I want to - I cant play games on it at the same time it is transcoding 4k content and it would be too expesive (electricity) to leave it on 24/7 anyway.
My smaller plex server is currently 3 days into āconvertingā a bunch of 4k content and is only 2% of the way through⦠THe Gaming PC would be finished by now.
I want to be able to offload cpu / gpu intensive tasks to the gaming PC while it is available.
All the plex server needs to do is check a āTranscoding poolā to see if any other, more powerful machines are available and send the jobs to those.
Itās old technology that is has been one on other platforms before and it would be excellent and solve a lot of scalability problems for us in the new 4K paradigm.
Please can we divert some of the resources going to DVR and 3rd party premium streaming services back into the core product?
I have a micro-cluster with 64 cores and several GPU-assisted workstations and it would be nice to be able to transcode my entire library using my cluster, especially during the initial setup.
All of my systems are also on a high speed interconnect (100 Gbps 4X EDR IB).
This would greatly help in my scenario. Iāve got 9 active transcode streams, and 4 direct, right now ā and my server can barely keep up. Granted, itās just an old workstation - but itās an i7-3770 with 16gb. Thatās not terrible.
Instead of having to constantly upgrade hardware and migrate Plex, because at this point it would need to be a new box, itād be great to have agents that can spin up on demand and off-load from the primary.
Not true. You can specify the listening port in the settings. I run two servers, both externally accessible, on my network just fine. Just have to port forward at the router, or in this case, via a load balancer.
Virtualize. Even if the machine is dedicated to Plex, get the P2V tool and make a VM out of your existing box and store it on a disk somewhere, wipe the physical, install VMware/KVM/proxmox and import the VM you made earlier. From then on, upgrading hardware is as simple as installing the hyperviser and migrating the VM. Most hypevisers donāt even require you to power the VM off to do so.
Iām sorry but this seems like a completely wasted feature for the vast majority of users. Unless every source file you have on the server is in 4k then transcoding it with decent quality should be fine on any six core CPU. Considering that AMD is literally throwing those at you for free paired with dirt cheap motherboards which work just fine for a Plex server it seems utterly redundant to start talking about solutions using dedicated render machines.
For instance I run Plex on a Ryzen 1600. It can easily handle 3-4 concurrent streams as long as only one of them is transcoding 4k. A friend of mine uses an old i7-5820k he has no issues either although he is not on the 4k band wagon yet as his TV is just 1080p.
Well I run and maintain a number of servers just because of this issueā¦
Distributed Transcoding would make my life so much easier⦠and eliminate the duplication of information that I have to store and scrape for each of my serversā¦
Yes I guess I could by a power hungry machine to do it, and considering that the sleep function doesnāt work anymore wouldnāt help my electric bill anyā¦
But figure this, in a 24 hour weekday most users only use their PMS server 5 to 6 hours at most, that means itās running for 18 hours needlessly⦠little bit less on weekendsā¦
do the math it adds up fast⦠Just wish Plex would see it and fix both.