Book a call with the founders to talk about Rthmn
From WebSockets
to gRPC
How we achieved 70% payload reduction and 3x faster serialization by migrating our microservices to Protocol Buffers and gRPC streaming.
Binary protobuf encoding vs JSON text
Reduced cloud egress costs significantly
Faster serialization, less parsing overhead
From 130μs to 13μs per message
The New Architecture
Multiple services communicating via high-performance gRPC streams, with WebSocket only at the client boundary.
Ingests real-time prices, calculates box arrays, streams to all services via gRPC
Scans millions of fractal paths in real-time, detects patterns, generates signals
Receives gRPC streams from internal services, broadcasts to clients via WebSocket
The Transformation
From verbose JSON text to compact binary Protocol Buffers.
Performance Gains
Daily Impact at Scale
The Numbers
Why gRPC Won
Binary Serialization
Protobuf encodes data in compact binary format. No field names transmitted, just raw, typed data.
HTTP/2 Multiplexing
Single connection handles multiple streams. No head-of-line blocking.
Type Safety
Schema-defined contracts generate typed code across all languages automatically.
Native Streaming
Client-streaming RPCs built into the protocol. No custom implementation needed.
Backpressure
Flow control handles fast producers and slow consumers automatically.
Error Handling
Standardized status codes and rich metadata. No more custom error parsing.
The Numbers Don't Lie
At scale, the savings compound quickly.
| Metric | Daily | Monthly | Yearly |
|---|---|---|---|
| Messages Processed | 17.28M | 518M | 6.3B |
| Bandwidth Saved | 33.7 GB | 1.01 TB | 12.3 TB |
| CPU Time Saved | 33.7 min | 16.8 hrs | 205 hrs |
Built for Scale
The bigger we grow, the more gRPC wins. What started as an optimization became a foundation for handling hundreds of markets in real-time with sub-100ms latency.