Yes, that’s a different issue. The upload of an asset isn’t slow, it’s the mass import of multiple assets that is.
FileZilla will stream multiple files at once or send them in a single continuous stream, and it has a single connection to the server. In the case of Forge assets, it’s doing a separate connection for every file and uploading it, waiting for success confirmation then doing the next file. Don’t forget that for each file you upload to the server, the server itself then needs to upload it to the geo-distributed S3 storage before confirming reception. Although that’s faster than your upload since the two servers are in the same datacenter, it still requires one additional connection to be established.
How long would it take if you used Filezilla to upload one file, then disconnect, reconnect, upload the second one, disconnect, reconnect, etc… for all 72 files ? Probably more than the 10-20 seconds you mentioned.
So to summarize, the issue you’re describing (which is definitely an issue and which I plan on fixing very soon) is not about “asset upload is slow”, but rather on “no mass-upload optimizations”, which is why I said it’s unrelated to this issue.
This specific issue was that it would take 1 to 3 seconds to process on the server side a single upload, and the more assets you had, the more time it took to process it. Now, you can have 0 assets or a million assets, it will always take the same amount of time, less than 10 milliseconds to process it (add it to your assets library, create the parent folders if they don’t already exist, update your total used quota, which is what took seconds before). That’s independent of the time it takes to establish the connection and upload the actual file content.
Are you saying that gap will close significantly with new features?
Yes, it should become much faster. I have a bunch of optimizations planned to make it faster, there’s always going to be the cost of having to do the upload from the server to the S3 storage, but I have been working on ways to try and optimize that as well.