I recently commented on a post from Nathan Rose about Dataverse Low Code Plugins stating my assumptions that they won’t perform as well as standard plugins.
So was I right? Lets find out…
Here’s how various types of ‘automated’ event handler (server side logic running in foreground, triggered by a row being created) performed for me in some aritifical benchmarks. (I included my project DvJs in the comparison out of interest, which was my solution to low code plugins for demos etc before official LCPs were announced).
IMPORTANT: YMMV and performance isn’t everything! Don’t optimise for performance prematururely and without considering the other qualities of the different approaches. (Having said that, also don’t assume the low code approach will always be easier to maintain).
Test case 1 - no actions (empty plugins) - just to see the overhead the platform is adding to load and execute them.
Test case 2 - set name - set the primary name column of the row being/just created to a static value.
Both 100 iterations each with a warmup phase. The DotNetPlugin is set as the baseline and the ‘ratio’ result is the relative performance, where 1=same and <1 is better.
Benchmark results Conclusions from these simple tests - LCPs added up to 40% overhead vs a .net plugin.