The AppGW autoscaling SKU (v2) can scale up to 125 instances (see App GW limits).
Each instance can support at least 10 capacity units. The number of capacity unit per instance depends on the application processing requirement (I couldn't find any max value).
A single capacity unit is composed of 1 compute unit + 2500 persistent connections + 2,22 Mbps of throughput:
- Each compute unit supports ~50 connection requests/sec (10 with WAFv2)
- If any of the 3 above parameters (compute unit capacity, # connections, throughput) is exceeded, additional capacity units get triggered, as stated:
➡️ See Application Gateway high traffic volume support | Microsoft Learn for further recommandations (max instance count set to 125 + min instance count based on the average compute unit usage).
💡 For a deep dive on Application Gateway, check out this great AppGW MicroHack by David Santiago!
| Instance | (Min) capacity unit | Persistent connections | Throughput | Connection requests/sec (Standard v2) | Connection request/sec (WAF v2) |
|---|---|---|---|---|---|
| 1 | 10 | 25 000 | 22 Mbps | 500 | 100 |
| 25 | 250 | 625 000 | 555 Mbps | 12 500 | 2 500 |
| 50 | 500 | 1 250 000 | 1,11 Gbps | 25 000 | 5 000 |
| 75 | 750 | 1 875 000 | 1,665 Gbps | 37 500 | 7 500 |
| 100 | 1 000 | 2 500 000 | 2,22 Gbps | 50 000 | 10 000 |
| 125 | 1 250 | 3 125 000 | 2,775 Gbps | 62 500 | 12 500 |