Meta changed one character of code and freed up 15,000 servers a year — thanks to eBPF

blaizedsouza · x · 2026-09-06

• Meta's Strobelight continuously profiles every production host at near-zero overhead; that visibility surfaced a one-character code fix saving capacity of 15,000 servers per year.
• How eBPF works: tiny sandboxed programs loaded into the running kernel — no patch, no reboot, no downtime; a verifier checks every instruction so the kernel cannot panic.
• Use cases: Cloudflare's XDP drops 10M+ packets per second per core inside the NIC driver, mitigating multi-terabit DDoS floods; AWS ships eBPF-based Cilium as EKS's default network layer, pushing 30-40% more throughput.
• The author also open-sourced pktz, an eBPF-powered live traffic monitor that accounts every byte per process and per connection, no packet sampling required.

Original post →

More from Infra

• Meta researcher: horrifying infra, not talent, is why OSS can't catch closed frontier — A_K_Nain · 2026-10-23
• Jane Street and NVIDIA speed up Mamba-2 with custom cuTile kernels — blelbach · 2026-10-23
• LatentLoop: asymmetric compute allocation cuts looped-MoE training cost by 19% — pmttyji · 2026-10-23
• Same AI request emits 8.5x more CO2 in Texas than in Sweden — jlundberg · 2026-10-23
• SK Hynix Pulls Plug on Solidigm US IPO as Chip Stocks Slide, Shares Down ~30% This Quarter — rohanpaul_ai · 2026-10-23
• Designer mocks Nvidia: buy idle GPUs or lose access to new ones—definitely not a bubble — michalmalewicz · 2026-10-23

Infra channel →