Network Integration Guide › About Network Integration › Choosing a Deployment Architecture › Connected Inline with a Network: Active and Passive Modes › Socket Connections to a Hub versus Directly to Multiple Policy Engines

Socket Connections to a Hub versus Directly to Multiple Policy Engines

The NBA can output data through a socket connection to a CA DataMinder hub or direct to policy engines. But which method is better?

Output to a hub

This method can provide more effective fault tolerance. That is, if a policy engine becomes unavailable, the hub is able to distribute load more efficiently across the remaining policy engines.

The disadvantage is that this method requires an additional network hop, with captured data passing to the hub and then onto policy engines. Also, the hub captures whole objects before forwarding them to the policy engines, so communication between the NBA and policy engines is slower.

Output to policy engines

This method is more direct and can improve network performance. However, this method provides less effective fault tolerance: if a policy engine becomes unavailable, captured items are redirected to the specified standby policy engine (if defined in nbapolicy.xml), but without any attempt to balance load.

By default, this method only uses one policy engine per NBA CPU, although you can change this by setting the 'Connections per Network Analyzer' value on the CA DataMinder Network Console's 'Policy' page.