Is Amazon S3 intelligent tiering useful for Discourse?

I’m looking at setting up Amazon S3 to store files for the Discourse instance I manage and I’m wondering about their intelligent tiering option, moving objects to cheaper semi-archive storage when they haven’t been accessed for 30, then 90 days.

With it, objects stored un-accessed for >90 days are approximately a sixth of the cost at rest than objects accessed in the last 30 days. In theory this could make files in older, less visited topics significantly cheaper to store in S3.

Using intelligent tiering is negligibly cheap ($0.0025 per 1000 objects at time of writing) so that’s not a factor but I’m wondering if it will actually have any benefit in practice. Most notably I’m thinking of crawlers, particularly poorly written or malicious ones which might request the files with some regularity regardless of date information provided by the server.

If anyone is using S3 intelligent tiering with Discourse, what has your experience been of objects moving to and staying in cheaper storage? Does it have any significant negative impact on viewing topics with images that haven’t been accessed in over 90 days?

2 Likes