Deribit has deployed a new matching engine that dramatically reduces order processing latency, cutting median response times from 4.7 milliseconds to 76 microseconds and tail latencies from 124 milliseconds to 234 microseconds. The upgrade, which processed over 90% of order entry messages during its rollout period, marks a significant structural improvement to the derivatives exchange’s core trading infrastructure designed to handle high volumes with more predictable performance.
- Median latency improvement: 61× reduction (4.7 ms to 76 µs)
- P99 latency improvement: 531× reduction (124 ms to 234 µs)
- Peak processing capacity tested: 1.02 billion events over two days; 23,836 requests in a single second
- Performance under congestion: Median latency rises only 10% from quiet to busy conditions
The new system prioritizes determinism and consistency over raw speed. Order entry and market data now operate on dedicated paths with fewer processing steps, while risk checks have been restructured to remove bottlenecks from the critical path. During testing on August 24-25, when the exchange processed its heaviest traffic, the worst round-trip time during the busiest second was just under 0.8 milliseconds, with the backlog clearing within 2 milliseconds, a timeframe that previously handled a single order.
For retail traders using Deribit‘s website or mobile app, the upgrade requires no action and operates transparently. The exchange’s classic API remains fully supported for brokers and technology providers, while institutional market makers and latency-sensitive firms can access new binary interfaces for direct, lower-overhead order entry and market data connectivity. The primary benefit is not speed alone but consistency: reduced execution uncertainty, more stable performance under load, and a more predictable environment for automated trading strategies.
Deribit FZE operates under regulation by the Virtual Assets Regulatory Authority and plans to leverage the new engine’s throughput capacity to expand product listings and support previously impractical features such as broader cross-collateral adoption.
