Stephen Blum, CTO at PubNub, discusses the significant performance improvements achieved by rewriting their Presence Service in Rust, a programming language known for low-level optimizations and safe concurrency. The transition from Python to Rust has resulted in dramatic latency reductions—up to 80%—and improved consistency across regions, reducing latency spikes and enhancing real-time communication. The rewrite involved several phases, starting with Presence Read APIs, which saw latency drop from an average of one second to 200 milliseconds, and followed by bundling Write APIs, leading to further efficiency gains by reducing multiple database calls to single bundled transactions. The new system not only improves speed but also decreases global resource consumption, offering cost savings and reducing environmental impact. As the Rust-based service currently operates with only 20% of the original resources, further latency reductions are expected as more Rust versions of APIs are deployed. The ultimate goal is to have the entire Presence Service running in Rust, including the Presence Event pipeline, to continue optimizing latency and resource usage.