ShareChat’s Path to High-Performance NoSQL: Q&A with Geetish Nayak

ShareChat, India's leading social media platform, faced challenges in achieving microsecond P99 latency for 1.2 million operations per second to support 180 million monthly active users, necessitating a modernization of their NoSQL database. Geetish Nayak, Staff Engineer/Architect at ShareChat, spearheaded a migration from their existing database to ScyllaDB, resulting in a 3-5x performance improvement and a 50-80% reduction in costs. The transition involved migrating 80TB of data and 40 services with zero downtime while adopting best practices such as ScyllaDB's shard-per-core architecture, which enhanced performance and reduced the need for additional caching solutions like Redis. The migration included deploying ScyllaDB across core use cases such as chat applications, real-time notifications, and data science feature stores. ShareChat's architecture relies on Apache Kafka for aggregating user interactions, which are then written to ScyllaDB using atomic counters to handle high-volume data efficiently.