Cluster Computing: A Case Study in Scaling a Fuzzy Search
by Riley Berton
Viggle is a mobile app and loyalty program for TV, that awards free points for engaging with your favorite television shows. These points are redeemable for real-world rewards. The main entry point to the mobile app is an audio verification of the television show you are watching.
The main focus of the talk will be the usage of a cluster compute approach for real-time fuzzy matching at a large scale as used in the audio verification system. I will address the following key questions:
- How do we deliver sub-second searches to tens of thousands of simultaneous users?
- What did we try first and fail at miserably?
- Why did initial approaches perform so poorly?
- How do we deal with instantaneous load as happens nightly during prime time television (the definition of Surge!)?
- What useful tips do I have when approaching similar problems?
- And, can I get some free points for watching this talk?