Bandwidth Planning for CI/CD Pipelines

CI/CD Pipeline Bandwidth Planning

CI/CD pipelines are surprisingly bandwidth-intensive. Each build may download dependencies, pull Docker images, run tests with external services, and upload artifacts. Understanding these flows helps size runner infrastructure.

Typical Pipeline Bandwidth Profile

A Node.js web application build:

Stage	Data Transfer
Git clone	50-500 MB
npm install (cold cache)	200-800 MB
Docker base image pull	100-400 MB
Test fixtures/data	10-100 MB
Build artifact upload	50-200 MB
Docker image push	100-500 MB
Total per build	500 MB - 2.5 GB

Calculating Required Bandwidth

For a team running 200 builds/day with average 1 GB per build:

Daily transfer: 200 * 1 GB = 200 GB
In 8-hour work window: 200 GB / (8 * 3600) seconds
= 200 * 8 Gb / 28,800 = ~55.6 Mbps sustained

Peak (10 concurrent builds): ~556 Mbps burst

Caching Strategies to Reduce Bandwidth

Strategy	Bandwidth Savings
npm/pip/Maven cache	60-90% of dependency downloads
Docker layer cache	70-95% of image pulls
Git shallow clone	80-95% of clone size
Artifact cache	50-80% of rebuild data

With effective caching, the 1 GB per build drops to ~200-400 MB.

Self-Hosted vs Cloud Runners

• Cloud runners (GitHub Actions, GitLab SaaS): Bandwidth is included but cold caches are common
• Self-hosted runners: Full control over caching, but you pay for bandwidth
• Hybrid: Use cloud for burst, self-hosted for frequent builds

Cost Impact

At AWS rates ($0.09/GB egress):

200 builds/day * 1 GB * 30 days = 6 TB/month
Cost: ~$540/month in bandwidth alone
With caching: ~$108-216/month