Virtual Machine / NAS Drive

[graham728]

Hi,

Has anyone installed this on a virtual machine on a NAS drive? We have the below NAS drive:

RS18017xs+
INTEL Xeon D-1531
2.2 GHz
Cores: 6
RAM: 16 GB

I was going to test it out with 2 cores / 6GB RAM and see how I get on. I'm not massively bothered about transcode times as it would be very short files being processed. Before I test it out, does anyone know if it's even possible? It would be a Virtual Windows 10 it would be running off.

Thanks.

[admin]

Hi graham728,

I don't see why virtual machine should not work. There MIGHT be som graphic issues but generally FFAStrans is very system friendly and Windows version agnostic as long as you are above XP and server 2003. Just pump up the CPU

-steinar

[ddehoff]

I am also looking to install this on a virtual machine. Can FFAStrans make use of multiple cores? How much RAM should I be looking at?

[emcodem]

Hey ddehoff


Well i assume you want to process non-professional videos with your ffastrans installation? The answer would dramatically change for audio only processing or professional Formats especially UHD and HFR stuff.

First of all, you should be aware about the tools underneath ffastrans. FFASTrans itself is more or less just orchestrating them, therefore it does not need multicore support or lots of memory (1GB should be totally fine). Ffmpeg is used for decoding/encoding, avisynth for filtering. You find lots of information on google about them.

For Video Decoding/Encoding depends a lot on the Formats you are decoding/encoding and your hardware setup. E.g. if you have a processor supporting intel quicksync, and you manage to pass the quicksync engine to your VM Guest, you typically need only about 2-4 Cores for the maximum speed in case your input/output format is quicksync compatible.
Anyway, in general the more Cores and the higher the frequency, the faster your processing will run. Most codecs scales up to a maximum of 32 Cores, windows does not support running one application with more than 32 Cores AFAIK.

Same with RAM usage. If you do UHD, 4:4:4 Colorspace, you will need around 16-32GB minimum but for a simple HD 4:2:0 4-8GB is absolutely sufficient.

For a better answer, you would need to provide more informations about what you are doing.

Cheers,
emcodem

[ddehoff]

Thank you for the response!

This will be for accepting a wide range of input formats uploaded by content creators. Some may be UHD or HFR but the output will only be HD in a standard codec, file format and dimensions. Currently the volume is low, perhaps 2-4 videos uploaded each day. I'm just looking to reduce the encode time that each user has to wait for each uploaded video.

My goal is to eventually have a farm of at least two Windows 2016 VMware VMs. I'm assuming that FFAStrans will have no problem on Windows 2016?

So you are saying that intel quicksync will improve the encoding speed? I will get with my IT folks to see if we can support that.

[emcodem]

Sure, Server 2016 is fine.
The following could be more detailed if you write the Codecs you are using (for encoding)
From your description, you should be fine with 8GB Memory. If you have enough resources, better go with 12 or 16GB.
Quicksync (same as Nvidia NVENC -> aka. "GPU" encoding) can only help you when producing distribution formats, e.g. high compressed H.264 or H.265 with low Bitrates (around 16 Mbit/s or so).

How to get best speed for a single transcode: If the input Files are using High Bitrate production codecs > 20Mbit, you'll need to get the fastest CPU with the most Cores that you can (typically Core i perform better than Xeon due to higher Clock speed). Sure this can be mixed, decode professional formats on CPU and encode low bitrate formats using GPU will deliver best Performance. On the other hand, decode high bitrate and encode high bitrate will bottleneck your CPU as both operations could make use 100% of the CPU.

However, back to the topic, When using a VM Guest for Video related operation, you will have a hard time getting the GPU Encoding Engine mapped into the Guest and also you will suffer from CPU power because VM hosts typically have many cores but very low Clock speeds.
In high volume environments, we do not virtualize encoder Servers as they constantly use 100% of the available CPU and or GPU resources so i would need one Server per VM Guest which makes the Virtualisation unneccessary - and i don't need to engineer around the topic how to get the precious GPU encoders mapped into the VM Guest.

cheers,
emcodem