• Rebuilt Datadog Lambda Extension in Rust, slashing cold start latency by 82%. • Reduced memory footprint by 40%, cutting usage from 55 MB to just 7 MB. • Shifted from heavyweight Agent to lightweight sidecar, collecting logs, metrics, traces, and profiles with minimal overhead. • Leveraged Rust’s zero‑cost abstractions and static linking to eliminate unused dependencies and compress binaries. • Implemented asynchronous telemetry aggregation, reducing CPU spikes and improving observability in resource‑constrained Lambda. • Achieved a new performance floor below 200 ms cold start, meeting AWS Lambda Extension benchmarks.

Article Summaries:

  • Datadog announced a rewrite of its AWS Lambda Extension in Rust to address performance and resource constraints in serverless environments. The new implementation cuts cold‑start latency by 82 %, reduces memory usage by 40 %, and shrinks the binary from 55 MB to 7 MB. Previously built atop the larger Datadog Agent, the extension struggled to meet the tight CPU and memory limits of Lambda functions. By reengineering the sidecar in Rust-leveraging its low‑overhead runtime and efficient binary size-Datadog achieved a significant performance floor reduction while maintaining full telemetry collection. The move underscores the trade‑offs and risks of rewriting production software for specialized workloads.

Sources: