Finish scans before your coffee cools ☕️

Stella Ops completes a warm SBOM‑first scan in ≈ 5 s on a 4‑vCPU runner and keeps cold‑path “first run” images below 30 s. Less waiting, more shipping.

How we squeeze the seconds out

SBOM‑first fast‑path

If your build already emits an SPDX or CycloneDX SBOM, Stella Ops skips layer unpacking and jumps straight to CVE correlation.

Δ‑SBOM diff engine

Only what changed gets re‑analysed. A 10‑layer image with one updated library scans in ≈ 1 s.

Hot cache

Redis LRU keeps the last 2 048 SBOMs in RAM — 0.3 ms look‑up for repeat builds.

Parallel walkers

.NET 10 LTS async pipelines overlap layer de‑compression with vuln‑look‑ups.

Real‑world benchmarks (4 vCPU / 8 GiB VM)

Image (public)ScenarioStella Ops v0.1‑αTrivy 0.51
nginx:1.26‑alpinewarm SBOM4.7 s12.4 s
python:3.12‑slimcold (Δ‑SBOM off)18.9 s31.8 s
bank‑api:prod@sha256:…Δ‑SBOM (1‑layer delta)1.2 s15.0 s

Figures averaged over 20 runs; full methodology lives in the benchmarks repo.

5 s

Warm‑path average scan time

Do I need to ship an SBOM?

No — Stella Ops can generate one, but providing CycloneDX or SPDX during the build shaves 60 – 80 % off runtime.
Read more
No — Stella Ops can generate one, but providing CycloneDX or SPDX during the build shaves 60 – 80 % off runtime.

How does Δ‑SBOM stay accurate?

Each layer digest is hashed; if any layer differs, Stella Ops falls back to a full scan to prevent false‑negatives, then caches the fresh SBOM.
Read more
Each layer digest is hashed; if any layer differs, Stella Ops falls back to a full scan to prevent false‑negatives, then caches the fresh SBOM.
CI integration recipes   Explore all features   Air‑gap ready — see Offline Kit