Is there a way to calculate how much compute to provision for the same instance when switching from N1 to N2 or N2 to N2D (assuming the same workload)?
All I could find is general information like "N2 offers 20% more performance to N1" or "N2D offers 13% better performance to comparable N series and 39% better performance vs N1".
Say I am currently using 100 N1 vCPU, does that mean I can expect to provision 80 N2 vCPU or 61 N2D vCPU for the same general purpose workload?
As you're interested to know about 1vCPU N1 vs 1vCPU N2 vs 1vCPU N2D refer to this link
Below which might help you in understanding its concept :
· N1 is the first-generation machine which can provide up to 96 vCPUs and 624 GB memory and is powered by Intel Skylake CPU or one of its predecessors. Google recommends usage of its second-generation machines (N2, N2D and E2) while using general loads, but it is interesting to note that apart from the Accelerated-optimized machine family, N1 machines are the only ones that can be used to support GPU workloads with a selected NVIDIA GPUs. N1s also support TPUs in select zones.
· N2 is the second-generation machine powered by 2nd Generation Intel Xeon Scalable Processors (Cascade Lake) CPU platform with a base frequency of 2.8 GHz and a sustained all core turbo of 3.4 GHz. It offers up to 80 vCPUs and 640 GB memory. N2s provide more memory per vCPU as compared to N1s.
· N2D is the latest second-generation machine which became generally available earlier this year on 6th April 2020. They are powered by 2nd Generation AMD EPYC Rome CPU platform and have a capability to provide up to 224 vCPUs and 768 GB memory, the highest combination possible in the general purpose family. In the google announcement, the N2D-standard-32 (32 vCPUs) performed 39% better than N1-standard-32 when evaluated using Coremark and 13% savings over N1 family
