Media Server :: FMS Origin-Only Clustering Setup

I was successfully able to setup FMS edge-origin cluster but Origin-only clustering is the one we wanted for our live applications. The only place I found was 'Large Scale Deployment with FMIS'. I guess for Origin Only Clustering, we need to have our own caching system. What does get cached in a Live AV Chat? How do I propagate it to other Origins? What kind of plugins would I need for cache managing? What should the plugin do?

In the attached Origin Only Architecture, what does the Primary Origin do and Secondary Origin do? I thought Origin is the place where the application is actually running and Edges do proxy. Is it the same here in Origin-only clustering also? Are there any tools which can measure the lag in AV? (like performance measuring tools or load testing tools for FMS). What kind of software Load Balancer can we use for FMS clustering? (Like Nginx, Zeus or something?)