Forget the Cold Start! Experience the "Spawn" Start in Serverless Computing

Many researchers reported considerable delay up to a few seconds when invoking serverless functions for the first time. This phenomenon, which is known as a cold start, affects even more when users are running multiple different functions orchestrated in a workflow. However, in many cases users need instant spawn numerous functions, such as parallel loops. This paper introduces the spawn start and analyzes the behavior of three Function-as-a-Service (FaaS) providers AWS Lambda, Google Cloud Functions, and IBM Cloud Functions when running parallel loops, both as warm and cold starts. We conducted a series of experiments and observed several insights that are beneficial for the research community. Firstly, cold start on IBM, which is up to 2 seconds delay compared to the warm start, is negligible compared to the spawn start because the latter generates additional 20 seconds delay. Secondly, AWS Lambda may run the same function in parallel loops faster than the warm start. Finally, Google’s cold start is “warmer” than the warm start of the same function.