I want to use Spaces as a storage for a big amount of pictures. It will in the end all be managed by a python application, most likely with the help of Boto3.
In the meantime, I need people to have access to the objects to work and order & rename them. And so I need to have a “file system” representation of the objects stored in spaces. Hence the idea of using project like s3fs and yas3fs to “map” objects to a local file system. So far I’ve had success using yas3fs (s3fs just doesn’t work, see https://www.digitalocean.com/community/questions/problem-mounting-spaces-bucket-with-s3fs). I am using a droplet instance located in the same region as our Spaces, sfo2, to speed things up. While this seems acceptable speed wise (apparently it only takes a long time the first time one open a dir with lots of files, after that it’s fast), it’s nowhere as fast as when using the DigitalOcean platform to check Spaces bucket content.
How can I achieve such speed? Would using Boto3 and build a we representation be faster than fuse mounted fs? Would it’d be as fast as DO’s own web interface?