O’Reilly Radar > Jamglue’s Growth and Scaling

In their presentation the Jamglue co-founders broke their usage into four phases. In phase one they used S3 to store audio files. Users upload audio files to JamGlue’s servers and the files would be sent to S3 at a later time. They didn’t want unrestricted access to the audio files so authenticated, auto-expiring redirects would be sent to users via the Jamglue servers. In phase two they started using S3 to store their user avatars. The avatars are never on JamGlue’s servers.Jamglue’s confidence in S3 grew and they now store all static content there (phase 3). This includes their CSS, JavaScript, and SWFs. They didn’t get into specifics, but they called out that they have had issues when S3 is slow and these files fail to load. The final phase moved the mixing process to EC2. The audio files are still on S3. EC2 instances grab the files from their remix them into a new audio file and then push the updated files back to S3. Planned future work includes having the audio files go directly to S3 and to move more computational processes to EC2. They may also need to buy more of their own hardware to speed up the site.

O’Reilly Radar > Jamglue’s Growth and Scaling


About this entry