Hacker News new | ask | show | jobs
by austinheap 2811 days ago
> How I read this, whether correctly or not, was that it’s suggested you set up an entirely new server, new database, new domain, new storage (cluster?), etc instead of being able to grow out your one instance.

I could see that interpretation. Personally, I read that as "if you need to load balance transcoding, that code won't be done by the core team (right now)". You can already use whatever Postgres/Redis/nginx setup you want, cluster/load-balance those how you desire, etc. Load balancing the transcoding engine for new uploads is simply a different beast.

> So now any new users need to register and upload to server2.example.com because of “because decentralisation”

The only scenario in which load balancing transcoding changes anything is if 1) a user is uploading videos to a specific instance and 2) that instance is busy transcoding videos that were uploaded previously. So if a user uploads a video to an instance where all the transcoding threads are already busy, it wouldn't be available until a thread freed up to transcode it.

> Please do correct me if I’m interpreting incorrectly

FWIW I'm not trying to put words in Chocobozzz's mouth, just my interpretation.