We used a distribution site architecture.
As an example
a primary ADT ‘Receiving’ site (received the ADT messages from the ADT system)
The primary ‘Receiving’ sites had one process with a limit of 11 OB threads.
Each of the OB threads connected to another site – Called the ‘Delivery’ sites.
the Receiving Site’s only job was to apply any global IB Tps procs then raw route ALL messages to each OB thread (a Delivery site).
The ‘Delivery’ Sites each typically had one process (some had more but not many more) and that process had a maximum of 11 OB threads which were typically destination systems. Here is where message filtering, transformation, etc. would take place.
So as I recall we had slightly more processes than sites because some Delivery sites had more than one process. Maybe 320 processes.
The above is a simplistic view – we did have some specialty integrations which required their own somewhat different architecture but most everything was expressed in the Receive Site/Delivery Site Architecture.
Our system size was dictated more by the volume and size of messages we integrated (nearly 7 million per day in 2015 and some of them very large) than by the site architecture.
Having multiple Recover y DBs as a result of the multiple sites was a plus for us.
email: jim.kosloskey@jim-kosloskey.com 29+ years Cloverleaf, 59 years IT - old fart.