Lightning Talk: Millisecond Scale-to-Zero and the Quest to Never Pay for Idle Again
summary
Unikraft-built unikernels boot and sleep in true millisecond semantics: builds from Dockerfiles produce OCI images via a craft CLI, cold-booting standard apps in ~26 ms, waking in about 7–8 ms, and end-to-end request processing around 60–70 ms. Scale-to-zero happens automatically; 1,000 Next.js instances wake five at a time during the live test and tens of thousands of Nginx instances run on one server (claiming 90,000 VMs in stress runs) using snapshotting to accelerate heavy apps.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 0:21Cloud is inefficient
- 0:56The problem with ScaletoZero
- 1:59Autoscaling
- 2:54Cold starts
- 3:39Cloud stack
- 5:03Unic kernel
related talks
24mCut your Cloud Computing Costs by Half with Unikraft
07mLightning Talk: Love at First Sight, Unik & Serverless
24mUsing Unikernels to deploy real-world apps
44mUnikernels and Hyper Elastic Clouds
38mUnikernels: The Next Evolution in Cloud Infrastructure
34mLook ma, no OS! Unikernels and their applications
09mLessons Learned from Building Fast Infra for Agents to Access the Internet
27mServerless MicroVMs for Containerized Applications