Adrian Colyer reviews an interesting paper:
Today’s paper analyses serverless workloads on Azure (the characterisation of those workloads is interesting in its own right), where users want fast function start times (avoiding cold starts), and the cloud provider wants to minimise resources consumed (costs). With fine-grained, usage based billing, resources used to keep function execution environments warm eat directly into margins. The authors demonstrate a policy combining keep-alive times for function runtimes with pre-warming, that dominates the currently popular strategy of simply keeping a function execution environment around for a period of time after it has been used, in the hope that it will be re-used. Using this policy, users see much fewer cold starts, while the cloud provider uses fewer resources. It’s the difference between the red (state-of-the-practice) and green (this paper) policies in the chart below. Win-win!
Very interesting.