Very interesting. But what impact did this have on the user experience? While the data makes sense as presented (i.e., aggregated) certainly there are some users outside the assumptions. What if they are mid to power users? Have they been upset?
As a feat of engineering...kudos. But engineering savings might not translate into user benefit. It would be interesting to see analysis from the customers' POV.
The post goes into detail about access frequency and migration between storage classes.
Their point is that storing older objects on IA saves money in exchange for latency. A user who tries to fetch a very old project out-of-the-blue can always live with very minor initial delay, after which it will stay on the instant access tier until timing out again.
In my case as a daily user, it’s zero. I pull up older (90+days) projects to reuse assets all the time and I never notice. As an AWS admin (past) I can definitely tell when other apps pull data from slower buckets.
Looks like you made this name 18 hours ago, at the exact same time that the person that posted this made their name and both this post and submission are your only activity. What is nice about it?
As a feat of engineering...kudos. But engineering savings might not translate into user benefit. It would be interesting to see analysis from the customers' POV.