Hi All,
I've been thinking about ways to build a really robust site spread across multiple cloud servers (common DB server) for redundancy and performance reasons.
I had an idea and I'm not sure if it's ridiculous or not:
1) create a multisite instance where all users, sessions and nodes are shared.
2) host each subsite on its own server
3) set up the subsites based on task, so that heavy processing can be done (many modules say) in one part of the site, and have less active modules/usage on another
4) use DNS or load balancer to direct traffic to the subsites based on task. So customised newsletters could all be generated on newsletters.site.com and all the video encoding could be done on encoder.site.com etc etc.
My thinking is that doing things this way allows you to run a minimal set of modules on the 'main' site, and also allows you to scale for each task individually, so if chat.site.com is taking an inordinate amount of resources, we can scale that server only without worrying about the rest.
Also if one function goes down, it doesn't take the whole site with it.
What do you guys think? Am I reinventing the wheel here or could this be a valuable approach?
BC
