# network ports needed for a two-node active passive cluster

We have a two-node cluster and each node has two network ports available. Before we move to the cluster, the way we use the current two network ports on our current production system is that one for public access and one for transactional push replication (replication between publisher and subscriber) and since there is no cluster involved, this network setup is fine. Now we are going to implement the cluster, it also involves a network for heartbeat between the two nodes so we have three network connection to be concerned and we only have two network switches (public and private). Since there are only two network port in a node, there are a few implementation I can choose from.
1. Public uses a network port, and replication and heartbeat share the other.
2. Public and replication use a network port, and heartbeat uses the other one.
3. install a third network card to each node so each has its own network port. Under this option, there will be two variations.
3.1 Put the public network and replication network to the publick switch, and the heartbeat network to the private switch
3.2 Put the public network to the public switch, and the replication and heartbeat network to the private switch
Option 3 is ideal but my question is that since heartbeat network utilizes very low bandwidth of the network. Is it a good idea of sharing it with replication network traffic as in Option 1 above? If a third network card is prefered, does it matter which option I use between 3.1 and 3.2?
Any pointer on this is very much appreciated.
wingman

I would go for option 3.1 . This is because I dont want to mix replication and heartbeart network traffic together. If i do, we may come occur an issue like failover due to just network congestion.