Hacker News Reader

>It's probably because playing rarely-played videos costs more in bandwidth.

Huh? Why would that be the case?

They exist on fewer servers, so they have to be pulled from deeper levels of cache, or maybe even no cache. Popular videos will be cached on edge servers all over the place, and will be quick and easy to deliver.

That's a good point that I didn't consider. Still, I don't think this was the right direction to go. If the videos were already rarely watched, going out of your way to make them less watched isn't going to save much bandwidth.

How do you know?

Azure, AWS, Google cloud all pay less for data that gets accessed regularly.

Google Cloud charges $0.02 / month for each gigabyte of storage, but if you store it in an infrequently accessed volume you can qualify for special pricing of $0.0012 / GB / month. That's a 177% difference in storage costs for cold storage vs hot storage.

So yeah, it's MUCH cheaper to store it in infrequently accessed volumes. And stopping people from finding the oldest videos allows them to do that.

According to Google Cloud, archival storage is only meant to be accessed once a year at most. Even infrequently watched videos should be considered hot (i.e. likely to be accessed more than once a month).

Reply on news.ycombinator.com