Cosmos DB VLDB paper shows better scheduling can be worth $100M+ a year
blaizedsouza · x · 2026-09-19
TiKV author siddontang highlighted a VLDB paper from Microsoft's Azure Cosmos DB team: "A Resource-centric Analysis and Optimization of NoSQL Workloads using Distressed Resource Volume Metric."
Key contributions:
- Open-source NoSQL workload traces from real Cosmos DB clusters
- A new reliability metric, Distressed Resource Volume (DRV), capturing user-experienced QoS
- LoadStar, an open-source policy simulation framework with a non-parametric QoS model
- Luna for forecasting load distributions and Orbit, a packing-and-migration algorithm for replica placement and rebalancing
siddontang notes the pipeline (production data → learn → decide → simulate → deploy) mirrors what they're building for TiKV's new auto-scaler, and asks whether PD scheduling and auto-scaling should eventually share one workload model — estimating better scheduling is worth $100M+/year: "scheduling is money."
More from Infra
- Open models now outspend OpenAI on Vercel gateway, taking 78% of token volume — soumitrashukla9 · 2026-09-19
- Multi-GPU rig cooling debate: should fan curves be tuned per GPU under load — TheZachMueller · 2026-09-19
- ROCmFix: zero-dependency Python tool auto-fixes AMD ROCm setups and benches Vulkan vs HIP — pavlexlkl · 2026-09-19
- Emulating memory access: FEX-Emu devs on the x86-to-ARM memory model minefield — blaizedsouza · 2026-09-19
- Three lines of bash escape Docker's Mac hypervisor: CVE-2026-77179 patched in Docker Desktop 4.88.0 — blaizedsouza · 2026-09-19
- Meta open-sources spmd_types: a type system to verify distributed PyTorch training correctness — austinvhuang · 2026-09-19