Date: 2026-01-31 Instance Type: m7i.large Technique: Cold launch across different AWS regions
| Region | API | Pending→Running | Boot→SSH | Total |
|---|---|---|---|---|
| us-east-2 (Ohio) | 2474ms | 16511ms | 2971ms | 21956ms |
| us-west-2 (Oregon) | 3827ms | 16375ms | 2632ms | 22834ms |
| us-east-1 (N. Virginia) | 3459ms | 16580ms | 3228ms | 23267ms |
| ap-northeast-1 (Tokyo) | 2654ms | 16679ms | 3996ms | 23329ms |
| eu-west-1 (Ireland) | 2544ms | 17065ms | 4463ms | 24072ms |
Fastest: us-east-2 (21956ms) Slowest: eu-west-1 (24072ms) Variance: 2116ms (9%)
Region has minimal impact on the core VM startup time.
Pending→Running is consistent across all regions:
- Range: 16.4s - 16.9s
- Variance: ~460ms (within measurement noise)
The total variance (1.5s) comes from:
- API latency (network distance to region)
- Boot→SSH latency (network distance for SSH)
- Pending→Running is region-independent - ~16.5s everywhere
- Closer regions have lower total time - Due to API/SSH network latency
- us-east-2 (Ohio) was fastest - Good balance of capacity + network
- Tokyo/Ireland slower - Network latency adds ~1.5s
- Region selection won't speed up VM startup - Only affects network latency
- Pick region for user proximity - Reduces API and SSH latency
- The 16s floor is global - AWS infrastructure overhead is consistent worldwide
- Multi-region warm pools - Would need pools in each region for latency
The Boot→SSH variance (2.7s - 4.1s) correlates with physical distance:
- us-west-2: 2.7s (closest to test location)
- us-east-2: 3.0s
- us-east-1: 3.4s
- eu-west-1: 4.1s (transatlantic)
- ap-northeast-1: 4.1s (transpacific)
Region API Pending Boot Total
------------------- --- ------- ---- -----
us-east-2 2382ms 16789ms 2991ms 22162ms
us-west-2 3965ms 16406ms 2672ms 23043ms
us-east-1 3266ms 16803ms 3435ms 23504ms
eu-west-1 2755ms 16767ms 4111ms 23633ms
ap-northeast-1 2697ms 16868ms 4138ms 23703ms