Performance Issues with Beam Runners compared with Native Systems

[damccorm]

While doing a performance evaluation of Apache Beam with Spark Runner - I found that even for a simple word count problem on a text file – Beam with Spark runner was slower by a factor of 5 times as compared to Spark for a dataset as small as 14 GB.

You will find more details on this evaluation here - https://github.com/soumabrata-chakraborty/spark-vs-beam/blob/master/README.md

I also came across this analysis called **Quantitative Impact Evaluation of an Abstraction Layer for Data Stream Processing Systems (https://arxiv.org/pdf/1907.08302.pdf / https://ieeexplore.ieee.org/document/8884832)

According to it, the observation was that for most scenarios the slowdown was at least a factor of 3 with the worse case being a factor of 58!

While it is understood that an abstraction layer would come with some performance cost - the current performance cost seems to be very high.

Imported from Jira BEAM-9440. Original Jira may contain additional context.
Reported by: soumabrata.