Now that I've explained the the implications of running a Network Load Balancing cluster in unicast or in multicast mode, I'll explain the advantages and disadvantages of each mode.
Of all possible cluster
The obvious advantage is cost: There is no special hardware to buy, and you do not have to invest in a second NIC for each cluster node. Admittedly, considering the low cost of NICs, those savings might be negligible. The second advantage is that you don't have to worry about router compatibility issues. Nearly every router supports unicast, although some do not support multicast MAC addresses.
However, I think the disadvantages – and there are two big ones -- of operating in unicast mode with a single NIC outweigh the advantages.
Disadvantage #1 is that communications with other cluster nodes is usually impossible. Disadvantage #2 is that this type of configuration usually leads to poor performance. Why? Because a single NIC must handle communications for both the cluster, and for ordinary network traffic. A primary reason for implementing a Network Load Balancing cluster is scalability. The idea is that two servers perform better than one. But if you implement a cluster design that degrades performance, you might be defeating the purpose of creating the cluster in the first place.
Nodes with a single NIC operating in multicast mode
Another common configuration for Network Load Balancing clusters is to operate nodes with a single NIC in multicast mode. As with the first configuration, there is no special hardware to buy, and no need to buy a second NIC for each server. But unlike the unicast configuration, this configuration has the added benefit of permitting normal TCP/IP communications between cluster nodes.
However, there are disadvantages to operating single NIC cluster nodes in multicast mode. As I mentioned, some routers do not support multicast MAC addresses. If your routers do support multicast, this will not be an issue.
The other disadvantage relates to reduced performance. This configuration uses a single NIC to handle normal network communications as well as communications specifically related to the cluster. In high-demand environments, the NIC could become a bottleneck.
Nodes with two NICs operating in unicast mode
As mentioned earlier, installing two NICs in cluster nodes, and configuring the nodes to operate in unicast mode is usually the optimum configuration. This configuration not only allows cluster nodes to communicate with each other, it increases performance by diverting all cluster-related communications to a dedicated NIC. In addition, this configuration operates in unicast mode, so there are no router compatibility problems.
But no solution is perfect. Operating a dual NIC cluster node in unicast mode has at least one disadvantage: Each cluster node requires a second NIC. You must consider not only the cost of this second NIC, but the time it will take your staff to install and configure this additional NIC into each cluster node.
Nodes with two NICs operating in multicast mode
The last configuration -- dual NIC cluster nodes operating in multicast mode -- also allows ordinary communications between cluster nodes. This mode also benefits from the increased performance associated with diverting cluster related traffic to a dedicated NIC.
The disadvantages? Some routers do not support multicast MAC addresses, and that you have the expense of a second NIC for each cluster node.
About the author: Brien M. Posey, MCSE, is a Microsoft Most Valuable Professional for his work with Windows 2000 Server, Exchange Server and IIS. He has served as CIO for a nationwide chain of hospitals and was once in charge of IT security for Fort Knox. He writes regularly for SearchWinComputing.com and other TechTarget sites.
More information on this topic:
- Tip: Sharing IP address among nodes in cluster leads to network issues
- Topics: Network load balancing
- RSS: Sign up for our RSS feed to receive expert advice every day.
This was first published in November 2006