Talks & speakers
Talks and speakers
The Quest for Speed - Loveholidays' Journey to 50% Better P99 Times with Go
David Vella
Software Engineer · loveholidays
Real-world performance tuning in high-traffic Go applications: how loveholidays cut P99 response times by over 50% across millions of holiday search requests. Covered profiling with pprof and go tool trace, correlating trace output with live traffic to find concurrency issues; treating performance metrics as a shipped feature with regression detection; and common Go anti-patterns (allocations in tight loops, naive string concatenation, unbounded goroutines). In his words: "We treat performance metrics like any other critical feature. If it regresses, we fix it before shipping."
Swiss Maps in Go - A Faster, More Efficient Map Implementation
Bryan Boreham
Distinguished Engineer · Grafana Labs
Go 1.24's Swiss Maps, referencing the Swiss Table algorithm used in C++ (Abseil SwissTable). Benchmark data showing up to 60% faster lookups on large read-heavy maps and around 25% lower memory use; SIMD/AVX2 support in Go 1.24; compact key metadata reducing cache misses; Grafana Labs telemetry-processing benchmarks. In his words: "Under the hood, Swiss Maps leverage clever SIMD optimisations to process buckets in parallel. It's like a free performance boost if your CPU supports it!"
Two performance talks: one on tuning a high-traffic production Go service, the other on a change to Go's own map implementation in the language runtime.
Organised by Humand Talent Solutions.