I’m a software engineer in Sydney working in a trading firm — some of the problems I work on are trading risk systems, HFT market controls, trade reconciliation, low-latency pricing, and the platform and observability layers underneath them.

Most of what I write about here comes out of that work: concurrency, measuring performance, and the ways distributed systems get correctness wrong.

Correctness and production diagnostics

• Amazon Aurora DSQL and the perils of write skewIsolation levels, and the anomalies optimistic concurrency still lets through.
• Why you should start using property based testingFinding the edge cases you would never have written a test for.
• Measuring .NET Core: a five-part seriesCounters, custom providers, gcdump for memory leaks, and crash dumps.

Performance, measured

• Counting sheep with BenchmarkDotNetSetting up a benchmark that measures the thing you think it measures.
• When to consider using ValueTask over Task with async codeAllocation on hot paths - and when the extra complexity is not worth it.
• Performance with ref readonly returnsAvoiding defensive copies of large structs.
• How to throttle http calls on the client using a DelegatingHandlerGetting maximum throughput out of a rate-limited endpoint.

Concurrency and throughput

• ConcurrentDictionary internalsStriped locking and the real cost of a resize, read from the source.
• The problem with the BlockingCollectionWhat a blocked thread per consumer costs a throughput-bound system.
• Why concurrency with async beats parallelism for scalabilityThread-pool behaviour under sustained load.
• Why is async void return bad?How an unobserved exception takes down the process.
• What is AyncLocal in C#?Carrying ambient context across an async call chain.
• Rx as part of your standard toolkitComposing asynchronous event streams instead of hand-rolling the coordination.

Latest

• Dec 2024 Amazon Aurora DSQL and the perils of write skew
• Sep 2021 How to throttle http calls on the client using a DelegatingHandler

All 58 posts →