Engineering
March 12, 2024
8 min read
How we achieved 10M ops/sec with FastDB Engine
Arindam Majumder
CTO & Co-founder
The performance of a database is often limited by the overhead of traditional storage engines. At FastDB, we decided to rethink how data is structured and processed from the ground up.
The Bottleneck
Most databases spend a significant amount of time in context switching and disk I/O. By moving everything to an in-memory, RDMA-native architecture, we've eliminated these traditional bottlenecks.
Technical architecture diagram would go here
Our engine uses a latch-free B+ tree implementation that scales linearly with the number of CPU cores. This allows us to process millions of transactions per second on standard commodity hardware.