Spawn start effects on serverless functions

Citation Author(s):
Christian
Hollaus
University of Innsbruck
Sasko
Ristov
University of Unnsbruck
Submitted by:
Sashko Ristov
Last updated:
Tue, 04/12/2022 - 14:12
DOI:
10.21227/qdnc-7y79
Data Format:
License:
124 Views
Categories:
Keywords:
0
0 ratings - Please login to submit your rating.

Abstract 

Effects of the "spawn start" of the Monte Carlo serverless function that simulates Pi.

The functions are orchestrated as a workflow and executed with the xAFCL enactment engine (https://doi.org/10.1109/TSC.2021.3128137) on three regions (US, EU, Asia) of three cloud providers AWS Lambda, Google Cloud Functions, and IBM Cloud Functions.

100 instances of the Monte Carlo function are executed with "concurrency=30", which means that the first 30 functions are executed in parallel, while the rest 70 are invoked one after the other after some active function finishes.

 

Instructions: 

The zip file contains 9 files, for each region separately.

Runtime is measured from the function and does not contain the overhead.