Hi Wayne,

In our cluster environment we are currently having around 5 instances running on a single cluster node. You can always have multiple instances on cluster node if you don't have enough harware and other resources. But multiples instances on a single cluster node is not usually recommended. Each SQL Server instance has its own SQL Server resource group which contains dedicated network name, I.P Address, Disk resources and SQL Server services.

When you have multiple instances on a single cluster node, there will be a huge competition among the SQL Server instances for the system resources. In that case you must setup the maximum memory for each SQL Server instance. Usually its recommended to leave 2-3 GB of the total available memory for the O.S and the rest to be shared among the instances. Also, its not advisable to run any other applications on that cluster.

Here is a very good article on the memeory management if you have multiple instances.

http://searchsqlserver.techtarget.com/tip/0,289483,sid87_gci1299068_mem1,00.html

Another major problem with this multiple instances approach is that suppose if you have 5 instances on node A and 5 instances on node B and if node A fails, then all the 10 instance will be running on a single node in which case you will have a huge and drastic performance hit. So you should always design a cluster with failover scenario in mind. i.e if you have 32 GB of RAM on node A, then it should be shared amoung the 10 cluster instances and the O.S in case of a failover. Clearly, this is a very poor design. I am telling this only to give you a picture of the process.

The installation is pretty much similar to installing multiple instances on a stand alone system. But during the installation, you should have an unique and dedicated Virtual name(Network name), I.P Address and disk resources for each of the SQL Server instances. Also, all these resources will be part of the resource group and at any time only one SQL Server instance can own that resource group i.e those resources in the resource group cannot be shared among multiple instances.

Thank You,

Best Regards,

SQL Buddy.