16 hours ago
- GitHub's aggregated uptime appears terrible (e.g., 89.43% over 90 days), but this is misleading because it sums downtime across well-isolated services.
- Individual services like core Git operations have 98.98% uptime, which is still poor but not 'zero nines'.
- Uptime numbers fail to capture partial outages, such as slowdowns affecting only specific user regions.
- The aggregation method punishes good engineering (service isolation) by making overall uptime look worse than individual services' uptime.
- The author concludes GitHub's uptime is more like a D grade than an F, and there are better reasons to criticize the company.