AWS Case Study: iLike

Nat Brown of iLike.com tells us how they use Amazon Web Services to scale out their Facebook application which has had over 12 Million installs since its launch in May 2007 and has reached a peak of 900,000 daily active users.

What is your title and main job function at iLike? Nat Brown, Chief Technical Officer. As with any startup, my main job function is to do everything and anything that needs doing. But, I do focus primarily on engineering features, database and software architecture, scalability/deployment, and longer-term technical goals.

Please describe your company and environment you work in. iLike, based in Seattle, will be turning 2 in March ‘08, but it really grew from our original musician-oriented community, GarageBand.com. In May of ‘07 we launched a version of iLike on Facebook.

Some say we compete in the “Social Music Discovery” or “Social Music Network” category, but we consider our company and its strategy—helping artists connect with fans and fans connect with their favorite artists wherever they are on the web—to be a unique combination.

How have you incorporated Amazon Web Services as part of your architecture? What services are you using? To date we have used primarily Amazon S3 to provide extra storage and delivery capacity for iLike’s Facebook application, especially to provide a buffer when our internal data-center build-out can’t keep up with our user and feature growth. User-generated content (pictures, videos) on our artist pages, as well as artist-generated content (especially iCast video posts to their fans) can be very high-volume traffic, and we are able to shift their delivery from our servers to S3 dynamically based on our needs.

We also use S3 for inexpensive off-line backup storage as part of our disaster recovery planning. We keep weekly and monthly backups and recovery logs from our databases in a secure S3 bucket.

Why did you choose Amazon Web Services? S3 is a no-brainer if you need to scale out delivery of simple content or provide secure limitless storage. We chose S3 for expediency (we were up and running within a few hours) and usage-based pricing (instead of minimum commits and long-term contracts).

Also, the active open source community, growing commercial ecosystem of products & services around AWS, and the high-quality tools for working with AWS are not only great resources for us but they also indicate that the community will be around for the long-term and so we feel comfortable basing more of our architecture on AWS.

What were you using as a solution before implementing AWS? How was the transition process? Prior to launching on Facebook and introducing iCast video posts for artists we had only occasional mobbed content on GarageBand.com and iLike.com. When our monitors would report high-traffic for some content we would shift it to a single managed server outside our data-center with a different bandwidth billing requirement. This required too much manual intervention and even its price was too high and involved figuring out our monthly commit, which is always hard to figure out with content that may or may not generate user interest.

Shifting this content to AWS/S3 was trivial. It was simple to automate with the backend tools/scripts that the S3 open source community and Amazon provide. Oh yeah, and it was 20x cheaper and we paid for only what we used.

How has AWS helped your Facebook app succeed? To date AWS has provided an incredibly flexible and cheap buffer for us during periods of capacity build-out where our data-center and architecture were not ready to handle sudden enormous growth. Specifically AWS has really helped us keep our site “up” when the scale of users and our infrastructure capacity should have toppled. In May ‘07 when we launched our Facebook application we could not have kept up without AWS behind us.

Do you have any future plans to incorporate other AWS solutions? Or use Amazon EC2 or S3 in other ways? Oh yes! We have many very large-scale compute jobs that we have started to shift into EC2 and expect to complete over the coming year. As we grow through the next 10x multiple of users and information about their music tastes we will be tuning our architecture so that our data-center servers are focused on just what they do best and we shift our huge quantity of utility compute work into EC2 where we can get it done more cost-effectively.

Is there anything else you would like to add? If I were starting up another company today I would think hard about building something that could live almost entirely in the AWS cloud. The cost of doing a web-startup has dropped 100x in the past 10 years, and the right project can be another 10x cheaper to build to scale using AWS if you think it through carefully.

↑ Top