I will also stay far away from these services. As a child post points out, if you do set a cap, you run the risk of cutting off a legitimate spike in high quali

I will also stay far away from these services. As a child post points out, if you do set a cap, you run the risk of cutting off a legitimate spike in high quality traffic. How do you tell the system to shut off due to an error and stay on during a spike in normal traffic? You can’t, unless you have someone watching it at all hours. That’s why I prefer to just rent whole machines on AWS. If I accidentally have some code stuck in an infinite loop making some O(1) call, don’t charge me $10,000 for that when it costs you nothing. If it’s actually consuming electricity and significant resources, I’ll know quickly because my service will go down, as it should, not scale infinitely until my company is bankrupt.