The Impact of Data Transfer on Performance in Elixir's Task.async

In Elixir and the BEAM (Erlang Virtual Machine), processes enable easy and efficient parallelism. Using functions like `Task.async/1`, developers often parallelize tasks expecting performance improvements. However, apparent in a benchmark PragTob conducted using `Task.async/1`, it was found that the sequential version was faster than parallel execution across various payload sizes. This surprising result led to the discovery that transferring large amounts of data across processes can have a detrimental effect on performance. BEAM's 'shared nothing' architecture requires data to be copied entirely between processes. This additional work of copying, especially for large data structures, can outweigh the benefits of parallel processing due to increased CPU and memory usage. PragTob suggests limiting the data sent to processes and using more efficient data structures to mitigate this problem.