I have Amazon S3 as my file storage server and EC2 instances as my application logic server. Now my app needs to upload some files which need necessary processing. I can think of two ways to do this:
Upload the file directly from mobile devices and get the file name and location(url), then send the url to my backend. Backend get the file by URL and do its job.
Send the file to backend using a multipart form, backend accepts the file do its job and finally save the file to Amazon S3.
Which is the standard way? What are the reasons?
Sending the object direct to Amazon S3 will be more scalable, less error-prone and cheaper (since you need less web server capacity to handle the uploads). Send the corresponding information to a Simple Queueing Service (SQS) queue that the back-end service can monitor and process. That way, if your back-end is ever offline, the jobs will simply queue-up and will get processed when the server is running again. A good use of loose coupling.
A third option would be to send the file directly from your mobile to Amazon S3, using Metadata fields to identify originating user, and then configure the S3 bucket to trigger some code in AWS Lambda that can process the file. This could do all the processing, or could simply trigger a process on your web server. Again, this reduces load on the web server and would not require sending a message to trigger the processing.
