[redat00]

Yes the pub/sub model gets way faster than the Ansible push model, especially at scale : When you start having a few hundreds of servers, you'll see that you just spend so much time waiting for the playbook to run.

It's also heavily bound to network connectivity, if your server your reaching has 100ms of latency, every time Ansible does something, you will feel those 100ms.

In the case of Peekl, it's easy : The server will read the content from the local disk (depending on what roles you have, your groups, and also load the variables.) And simply sends the "compiled catalog" to your agent who will go through all of it and performs actions as needed.