The portion about the cache servers has no explanation other than that we would need a load balancer to our cache servers.

If there are cache servers, then the assumption must be that these servers are faster to get to somehow then going against the persistent store or the database. In that case what do we do to maintain cache coherency ? are the writes happening from the cache servers? It would be nice to flesh out this portion of the design instead of some high level diagram that plops these boxes called cache servers.