s3 and conversion server
Jamroom Developers
DannyA:I think there is also something to be said for auto-scaling. Being able to monitor capacity of existing transcoder cluster and spin up additional servers on demand can really help if you have very spikey encoding demands. No sense leaving servers running unnecessarily.
The way Jamroom does everything via queues, this is really not as imperative as it may seem - all that happens is your queue latency gets a bit higher when you're in a "spike". Since it's not customer facing (i.e. your users aren't waiting on it to get their page rendered) I wouldn't worry about it.
Quote:Is it safe to say that one worker queue job is allocated per core? Does that apply to any queue task?
That's controlled by you. What we've done for our transcoding service is set it up so if we have 8 cores (for example), we have 7 video workers and 4 audio workers. The audio workers are much "lighter" on the CPU and can squeeze in to that last core and the "free" time on the video workers CPU.
Hope this helps!