MongoDB was deployed at Castlight in late 2012 to solve a significant scaling problem involving the retrieval of large sets of healthcare prices from a very quickly growing database. The most frequent search query is based on the geo location of the user, but includes several other user-specific filters. The key parts of the solution were:
- Geo index
- Very low latency reads
- Replica sets
- Flexible schema
SSDs
The Geo haystack index in MongoDB turned out to be the perfect solution for our problem. A strategy of replica set flipping also enables Castlight to swap in very large changes to the pricing data with no impact to the running application. SSDs were used due to the need to retrieve large, scattered data sets very quickly.
