Use data streaming in audio encode process.

[aomi]

The current code does the following:

• Downloads S3 audio file to be concerted to /tmp (execution context)
• Encodes the audio file using a ffmpeg process which opens said file and dumps the converted file to disk.
• Uploads the converted file to S3.

To make this process more efficient (and therefore reduce our Lambda execution time = less money), we can take advantage of streaming data!

Instead of downloading the entire data, we can stream it into stdin for the ffmpeg process and stream out the data into S3. This way, we can maximize all aspects of the lambda execution!

Some useful keywords to Google stuff with.

• ffmpeg pipe.
• S3 Streaming Upload/Download
• Go io.Reader / io.Writer interface usage.

[aomi]

If you look at the timing, you can see the waste time spend and idle time doing nothing.

[aomi]

While done in theory, the actual benefits are hard to see but this is good enough for now.