3 ms·

I guess technically you could store original and encode to <whatever format is popular> on-demand then just cache the most popular ones

by ilyt 3y ago

PaulHoule 3y ago
In internet-scale systems complex delivery systems involving transformation and caching are not necessarily a win for performance insofar as once you give up a millisecond of latency you cannot get it back. You very much can get a win out of a CDN but somebody has to do aggressive tuning of absolutely everything.
- ilyt 3y ago
  It's not for performance but for not having to permanently store terabytes of differently encoded data. You could possibly also pre-populate cache - all new media files get sent to encoder and cached preemptively, and if they are popular enough they just stay there till they are not.
andrewaylett 3y ago
Yes, we do that. Well, the CDN does the caching -- but we still get a 99.99% hit rate. Please forgive the case study format, but I'm quite proud of what we did and this is the easiest way to back up my assertion with data: https://aws.amazon.com/solutions/case-studies/skyscanner-cloudfront-case-study/ https://aws.amazon.com/solutions/case-studies/skyscanner-clo... We're currently generating JPG, WebP, and PNG, depending on the source and the target. The problem with doing this on-demand is that formats that are great for delivery but slow to encode aren't useful. I did try to add AVIF, but the latency on the first request was too high for it to be practical.
- ilyt 3y ago
  I guess "preheating" cache with preemptively encoding and putting new content in cache before it is shown to client could work? But still, anything AV1 based just needs a lot performance to encode that just doesn't feel worth it at the moment. Our system to encode videos was tooled for that but developers eventually decided it's too slow/cpu hungry to bother. It was some time ago tho, encoders did get better...

I guess technically you could store original and encode to <whatever format is popular> on-demand then just cache the most popular ones

by ilyt 3y ago

I guess technically you could store original and encode to <whatever format is popular> on-demand then just cache the most popular ones

PaulHoule 3y ago

In internet-scale systems complex delivery systems involving transformation and caching are not necessarily a win for performance insofar as once you give up a millisecond of latency you cannot get it back. You very much can get a win out of a CDN but somebody has to do aggressive tuning of absolutely everything.

ilyt 3y ago
It's not for performance but for not having to permanently store terabytes of differently encoded data. You could possibly also pre-populate cache - all new media files get sent to encoder and cached preemptively, and if they are popular enough they just stay there till they are not.

andrewaylett 3y ago

Yes, we do that. Well, the CDN does the caching -- but we still get a 99.99% hit rate. Please forgive the case study format, but I'm quite proud of what we did and this is the easiest way to back up my assertion with data: https://aws.amazon.com/solutions/case-studies/skyscanner-cloudfront-case-study/ https://aws.amazon.com/solutions/case-studies/skyscanner-clo... We're currently generating JPG, WebP, and PNG, depending on the source and the target. The problem with doing this on-demand is that formats that are great for delivery but slow to encode aren't useful. I did try to add AVIF, but the latency on the first request was too high for it to be practical.

ilyt 3y ago
I guess "preheating" cache with preemptively encoding and putting new content in cache before it is shown to client could work? But still, anything AV1 based just needs a lot performance to encode that just doesn't feel worth it at the moment. Our system to encode videos was tooled for that but developers eventually decided it's too slow/cpu hungry to bother. It was some time ago tho, encoders did get better...