Simon
679 posts

Simon
@simonjzhou
sunflower capital | prev: @unusual_vc @mckinsey @ucla | views expressed here are my own






Can you 4x OTel thoughput with Rust? Meet Rotel: pushing 460K spans/sec/core (vs 130k) The @streamfold team told us how it works -> clickhou.se/4rju9en ClickHouse is fast at ingesting OpenTelemetry data - so fast that ingestion layers sometimes struggle to keep up. The OTel Collector can be resource-intensive, and at petabyte scale, every CPU cycle counts. Rotel is a lightweight OpenTelemetry collector implemented in Rust that can 4x throughput vs the standard OTel Collector, while dramatically reducing CPU overhead. How? By diving deep into Rust's threading model, resolving allocator lock contention, and reengineering JSON serialization with RowBinary encoding to stream data efficiently into ClickHouse. At 460K trace spans/sec per core, Rotel makes OTel as efficient as ClickHouse.



The most important conversations today happen over text. The $10M close. The GP intro waiting on read. The colleague checking on dinner. All buried in the same inbox. Sometimes you blink and you miss it. Instant messaging is a privilege, not an entitlement. We've spent years clawing our way out of email, only to rebuild the same prison with a faster notification system. Until now. RPLY on your own terms. Available on macOS (and iOS). Stay dialed. @heynox













