Montse Medina, COO, Jetlore
I will describe the challenges we faced when designing a MongoDB database for processing large data streams and the solutions we applied. Some of the difficulties included write-intensive loads, uneven access patterns (posts with many followers get many more hits than posts with few followers), and non-trivial support of privacy. I will describe the choices we made for schema design to optimize writes and efficient querying/retrieval. I will also talk about indexing strategies, tradeoffs we made to work around MongoDB design, and reasoning we applied to find the most optimal denormalization of collections.