PERFORMANCE Signal 124
Jane Street podcast episode 18 discusses why trading-system optimization is harder than hyperscale
↗
Jane Street's Signals and Threads podcast episode 18 features Andrew Hunter discussing performance engineering, contrasting hyperscale optimization with the trickier bursty, low-latency workloads of trading systems.
For engineers optimizing systems, the episode highlights that scale changes the optimization calculus: at hyperscale, tiny gains have huge impact, but trading systems require different techniques due to bursty, low-latency demands. It also covers profiling tools and the differences between optimizing OCaml and C++, which are directly relevant to performance work.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Andrew Hunter worked at Google for seven years on multithreaded architecture and was tech lead for tcmalloc before joining Jane Street.
The episode argues that finding performance wins in trading systems is often trickier than at hyperscale because of bursty, low-latency workloads.
The discussion covers Andrew's favorite profiling techniques, visualization tools, and the unique challenges of optimizing OCaml versus C++.
THE CLUSTER