At least it should be checked how OpenTripPlanner's performance differs on different virtual machine types
I did load testing with D8 v3, H8, E4 v3, DC4s, F8s v2 and our current vm type D3 v2. Out of the new ones, F8s v2 had the best performance/price ratio. Compared to our existing VMs, F8s v2 has twice as many vcpu (4->8) and 1/7th more ram (14->16gb). Price with normal pricing model is roughly 45% higher but if we use the 3 year discount it is almost half the price of the current vms (https://azure.microsoft.com/en-us/pricing/details/virtual-machines/linux/). According to azure benchmark scores (https://docs.microsoft.com/en-us/azure/virtual-machines/windows/compute-benchmark-scores) new VM should have roughly 70% better performance. I found in load testing results with jmeter using OTP routing and other queries, that the average response time was about 45% less than with new vm and throughput was about 45% better. Therefore, we should get over three times better performance for the same price with new vms and pricing model.