Case Study: Zillow scales background and real-time workloads with ScyllaDB

Zillow processes 6,500 records per second with ScyllaDB

Zillow, a major real estate website, faced a challenge in processing its vast property and listing data from multiple message queues that could not guarantee message ordering. This created a risk of data inconsistency for their consuming services. To solve this, their engineering team utilized ScyllaDB as a highly scalable storage layer for their listing processor service.

By implementing ScyllaDB and employing techniques like using write timestamps, Zillow solved the ordering issue without application locks or transactions. The solution from ScyllaDB supports both real-time user-facing workloads and massive background processing. The results are significant scalability, with the ability to auto-scale background processing to 35 instances handling over 6500 records per second, all while running on a highly efficient three-node ScyllaDB cluster.

Zillow