Cloud engineers shouldn't have to choose between fast, safe, and cheap.
For twenty years, infrastructure has forced a triangle: pick two. Containers boot fast but share kernels. VMs isolate but boot slow. Functions scale but fence in your stack. Unikraft Cloud is what happens when you refuse the trade-off.
Built for milliseconds, not minutes.
Three architectural rewrites — VM stack, network layer, app startup — engineered for one outcome: every workload starts, scales, and stops at hardware speed.
From dead-cold to first byte. 100× faster than typical alternatives.
Strongly isolated VMs on a single machine — Prisma runs PostgreSQL at this density.
Specialized images carry only what your app needs.
Two orders of magnitude — measured in production.
A standard Dockerfile.
A radically different runtime.
Your existing Dockerfile in. A specialized microVM image out — typically 10× smaller. Built on custom infrastructure tuned for millisecond cold boot times.
Dockerfile → microVM
Same source. Same registry workflow. A radically different runtime artifact comes out the other end.
FROM node:20-slim
WORKDIR /app
COPY . .
RUN npm ci --omit=dev
CMD ["node","server.js"]- 01 No Dockerfile changes — same files, same registry workflow
- 02 Built locally with the unikraft CLI, pushed as a standard OCI image
- 03 Direct push: each UKC host runs its own OCI registry — no internet hop
- 04 Hardware isolation per instance — every microVM is its own VM
Idle to first byte —
inside one RTT.
The proxy buffers the request. The controller resumes the right instance from snapshot. The VMM brings the microVM live. The app replies — all within a single round trip. Animated below, phase by phase.
Cold-start within an RTT
A request lands. The proxy holds it open. The controller resumes the right instance from snapshot. The microVM ignites. The app replies — all before a traditional cloud has finished pulling its image.