Using ping to test AWS VPC network latency across availability zones in each single region
Overview
The code for this experiment can be found here.
Earlier I wrote an article to test the AWS VPC network latency with ping. That article only covered the test in my default region, ap-northeast-1. So in this article, I am extending it, to run the same inter-availability-zone latency test, but for each of the regions where I have access.
Results
| source availability zone | target availability zone | avg(ms) | min(ms) | max(ms) |
|---|---|---|---|---|
| ap-northeast-1a | ap-northeast-1c | 2.786 | 2.613 | 5.116 |
| ap-northeast-1a | ap-northeast-1d | 2.014 | 1.818 | 4.743 |
| ap-northeast-1c | ap-northeast-1d | 1.065 | 0.977 | 1.163 |
| ap-northeast-2a | ap-northeast-2c | 1.334 | 1.178 | 1.454 |
| ap-south-1a | ap-south-1b | 1.353 | 0.834 | 12.508 |
| ap-southeast-1a | ap-southeast-1b | 1.2 | 1.081 | 2.404 |
| ap-southeast-1a | ap-southeast-1c | 0.729 | 0.628 | 1.324 |
| ap-southeast-1b | ap-southeast-1c | 1.406 | 1.314 | 2.048 |
| ap-southeast-2a | ap-southeast-2b | 1.34 | 1.237 | 1.757 |
| ap-southeast-2a | ap-southeast-2c | 1.144 | 1.037 | 2.228 |
| ap-southeast-2b | ap-southeast-2c | 1.599 | 1.401 | 4.699 |
| ca-central-1a | ca-central-1b | 0.665 | 0.581 | 0.942 |
| eu-central-1a | eu-central-1b | 0.954 | 0.884 | 1.089 |
| eu-central-1a | eu-central-1c | 1.128 | 0.922 | 3.781 |
| eu-central-1b | eu-central-1c | 0.731 | 0.628 | 1.14 |
| eu-west-1a | eu-west-1b | 0.765 | 0.684 | 0.951 |
| eu-west-1a | eu-west-1c | 0.945 | 0.839 | 1.456 |
| eu-west-1b | eu-west-1c | 0.766 | 0.613 | 1.987 |
| eu-west-2a | eu-west-2b | 1.119 | 1.033 | 1.678 |
| eu-west-2a | eu-west-2c | 0.774 | 0.651 | 2.182 |
| eu-west-2b | eu-west-2c | 1.26 | 1.198 | 1.545 |
| eu-west-3a | eu-west-3b | 1.015 | 0.87 | 1.679 |
| eu-west-3a | eu-west-3c | 1.118 | 1.045 | 1.193 |
| eu-west-3b | eu-west-3c | 1.08 | 0.987 | 2.114 |
| sa-east-1a | sa-east-1c | 1.257 | 1.133 | 2.643 |
| us-east-1a | us-east-1b | 0.807 | 0.733 | 0.9 |
| us-east-1a | us-east-1c | 0.962 | 0.912 | 1.037 |
| us-east-1a | us-east-1d | 1.259 | 1.021 | 4.742 |
| us-east-1a | us-east-1e | 0.876 | 0.604 | 6.27 |
| us-east-1a | us-east-1f | 1.143 | 1.006 | 2.409 |
| us-east-1b | us-east-1c | 0.79 | 0.729 | 0.903 |
| us-east-1b | us-east-1d | 0.835 | 0.748 | 1.204 |
| us-east-1b | us-east-1e | 0.952 | 0.805 | 2.747 |
| us-east-1b | us-east-1f | 1.022 | 0.936 | 2.073 |
| us-east-1c | us-east-1d | 0.657 | 0.59 | 0.749 |
| us-east-1c | us-east-1e | 0.972 | 0.91 | 1.065 |
| us-east-1c | us-east-1f | 0.863 | 0.7 | 1.575 |
| us-east-1d | us-east-1e | 1.196 | 0.984 | 4.261 |
| us-east-1d | us-east-1f | 0.969 | 0.875 | 1.52 |
| us-east-1e | us-east-1f | 0.971 | 0.808 | 2.316 |
| us-east-2a | us-east-2b | 1.006 | 0.934 | 1.116 |
| us-east-2a | us-east-2c | 1.287 | 1.209 | 1.682 |
| us-east-2b | us-east-2c | 0.642 | 0.496 | 1.699 |
| us-west-1a | us-west-1c | 1.364 | 1.276 | 1.91 |
| us-west-2a | us-west-2b | 0.895 | 0.83 | 1.144 |
| us-west-2a | us-west-2c | 0.939 | 0.542 | 6.021 |
| us-west-2b | us-west-2c | 1.34 | 1.165 | 2.62 |
Please be noted that I didn't test latency between two subnets in the same availability zone.
The previous article showed that the following two cases didn't produce any noticeable latency difference:
- ping from one subnet to another, within the same availability zone
- ping within the same subnet, within the same availability zone
So I am assuming these two cases are the same in terms of latency.