GCP + BuildKit

Cloud Build Docker Cache Workbench

Externalize layer caching to Artifact Registry with mode=max on ephemeral VMs
BuildKit Execution Simulator
CACHED
Total Duration 36s
Baseline (Cold) 51s
Seconds Saved 15s
Pkgs Downloaded 0
Build Strategy Active: Builder stage intermediate layers recovered from Artifact Registry :buildcache image via docker-container driver and mode=max
STAGE 1: BUILDER (uv:python3.13-bookworm-slim) CACHED
STAGE 2: RUNTIME (python:3.13-slim-bookworm) REBUILT
GCP Config & Artifact Generator

    
FinOps Monthly Savings
Build Mins Saved: 150 mins
Estimated Spend Saved: $2.40 / mo
GreenOps Sustainability
Avoided CPU Burn: 2.5 vCPU-hrs
Avoided CO2e: 0.88 kg CO2e
Based on 0.35kg CO2e / vCPU-hour datacenter baseline

Trace the saved cache fixtures and their explicit limits

Read the explanation

In the saved Cloud Build workbench, the standard cold baseline is assigned fifty-one seconds. BuildKit registry max with a code-only change subtracts fifteen to display thirty-six seconds, while preserving builder sync as cached and packages downloaded as zero. Bars show fifty-one and thirty-six assigned seconds at nine pixels per second. Source comments claim outside measurements, but this explainer only establishes the written fixture, not current cloud performance. The heavy dependency toggle adds sixty to the fifty-one-second baseline, yielding one hundred eleven. The max code-only branch also increases saved time by sixty, so its assigned warm result remains thirty-six seconds. Bars show those one hundred eleven and thirty-six fixture seconds at four pixels per second. A dependency bump instead invalidates sync and subtracts only three from baseline. These branch values are model assumptions, not measured builds. Monthly minutes saved are rounded seconds saved times the entered build count divided by sixty. The standard code-only max branch saves fifteen assigned seconds each: six hundred builds yield one hundred fifty minutes, while twelve hundred yield three hundred. Bars show those assigned monthly minutes at one point five pixels per minute. Cost, CPU and carbon labels are further source assumptions, not current prices, financial savings or emissions measurements.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.