0

I'm currently working on a solution to online process uploaded video data.
There will be diffeent video-formats uploaded to the platform and we need to transcode them into flv & re-render them in a smaller resolution. (^= Compression)

My current plan is:
1 Webserver managing the web-application
1 Sotrage-Server for uploading and video-storage purposes

The transcode-process will take place on the storage-server (which will be a high-end and performant one) AND - if neccessary it will outsource this video-processing to Amazon-EC2-instances.

So that's my plan. I'm a bit doubtful if this approach is efficient and well-concidered.

So - In plaintext: I want some feedback of you!

Hope, you can help me!

Regards,
~MxAgent

Micronax
  • 660
  • 13
  • 25

1 Answers1

0

Start with local encoding and then move to the cloud if and when you need to.

A video encoding instustry heavy height, Sorenson, has a cloud-based solution that you could use: Sorenson Squeeze Managed. A potentially nice aspect of this solution is that if your encoding needs ramped up to the point where it made sense to do them internally, then you can purchase they software to run on your hardware.

Sidebar: Modern i7+ CPUs absolutely rock at video encoding. Assuming a multithreaded encoder (eg: x264 + FFmpeg with ptheads, where 4 threads is optimal), and with many cores (say 4+) it makes sense to run mulitple jobs at once. I currently use an algorithm like this to determine the number of concurrent h264 jobs:

# encoding jobs = max(2, (# cores / 4) + 1)

I mention this because, depending on your app, you may not need nearly as much encoding power as you think. If you are looking to "roll your own", then definitely start small and add complexity as you need it.

Community
  • 1
  • 1
Stu Thompson
  • 38,370
  • 19
  • 110
  • 156
  • Really thank you for this extensive answer! I think this'll help me much - especially this alrogithm! We're talking about an amount at the beginning of up 150 Videos per day at about 2-10mins - BUT: This can grow exponential after a few months or even weeks. So I need to be well-prepared! I think I'll use our XEON Quadcore Server which supports Hyperthreading and later on move to the cloud - when the queue grows and grows.. – Micronax Nov 15 '11 at 16:04