We had this:
Pri-Hub (nas)
^^^^
4 Tunnel Servers
^^^^^^^
Client Tunnel Client Hubs
This was hitting the subscriber limit and we had to tell folks to log off the primary hub and use the tunnel servers when they needed to use IM. We have 4 get queues from the Primary to each Tunnel Server (Alarm, QoS, Audit, Discovery) but when you factor in Baseline & Prediction that's almost 6 queues to each tunnel so that's 24 links from Pri-Hub to each Tunnel + everyone logging into the Pri-Hub = Way over subscribed.
To quickly check select your Primary Hub probe and Hit Ctrl+P, then in drop down "list_subscribers" Scroll all the way to the bottom and that's the # of subscribers.
We recently re-structured the overall and threw in a middle hub between the Pri-Hub and the Tunnel Servers.
Pri-Hub (nas)
^
Hub Collector
^^^^
4 Tunnel Servers
^^^^^^^
Client Tunnel Client Hubs
Now our Primary only has the 4 get queues to the Middle Hub and then middle hub has all the gets to the tunnel servers. Its been running 100% better since. All middle hubs are CentOS and Primary is Windows.
We also only have the nas probe on the primary hub. Our subscriptions to the primary averages in the 40's when everyone is on.
We also hit a limitation on the tunnel servers. We expected that the Linux version of the hub would support many more client server hubs but no. We found that anything more than 30 would hit issues. Resource wise on the box was minimal bu the hub probe just could not keep up. We now limit each to 25 to be safe.