@scott We actually had a recent breaking change that we should have documented here in regards to the cloud. I’m going to remove the solution tag as it will depend on test size entirely now. We do plan to introduce completely unique VU IDs that would remove all this messy math. I am not sure what that timeline is, however.
That said, we’ve introduced some tiering of hardware for cloud tests to improve spin up times, data processing and general stability. This tiering doesn’t impact the resources per VU as we linearly increase instance size (I know you didn’t ask about this, but I’m sure someone will read this in the future and will have that question). Anyway, here is how it goes:
We have 3 tiers of hardware for load-generation. The tier we choose depends on the number of VUs allocated to a load zone.
Tier 1 is used when there are 1-999 VUs in a load zone
Tier 2 is used when there are 1000-4001 VUs in a load zone
Tier 3 is used when there are more than 4001 VUs in a load zone.
The tier 1 server handles up to 300VUs
The tier 2 server handles up to 1200VUs
The tier 3 server handles up to 5000VUs.
For example, if you start a test with 900VUs, we will use 3x Tier 1 servers.
If you start a test with 1000VUs in a single load zone, we will use 1x Tier 2 server. If the same test is started in 2 load zones, there will be 500VUs per load zone and 4x Tier 1 servers will be used.
So you will need to determine, based on test size and config, what size machines you will use, then you can use the correct max VUs per load gen.