Hello Heriyanto,
Were you able to come up with something to solve your problem? We don't currently have any sort of sample catalog that could be used to easily benchmark the performance difference in your environment, though your e-mail has prompted some internal discussion about how we could conceivably provide something like that in the future.
I suspect that if you simply time any agent run in your environment (for an agent with a non-trivial number of resources) against legacy storedconfigs, and then time it again with PuppetDB, you should notice some difference (because legacy storeconfigs must synchronously insert all of the catalog data into the database before the run completes). You could also run your master with the "--debug" flag, which might provide you with some timing info for the various phases of the process.
However, the biggest benefits come when you have a master that is under some load from concurrent agent requests. That might be a bit more tricky to simulate in your own environment if the blog posts and other data we've posted isn't sufficient, but we would be happy to try to help if you have questions!
Thanks
Chris