Considering the platform: 4-way, 16GB RAM running Windows 2003 (32-bit)
Enterrise Edition and SQL 2000 Enterprise Edition. With only 2GB (or
3GB with /3GB switch) of pagable memory to work with, how is this
memory shared across multiple SQL instances within a clustered
environment? This is a new cluster that will potentially support about
20 databases. Any suggestions on how to best work through the decision
process of whether to add additional SQL instances and what the memory
implications would be either way?

Thanks much,

Ed

If you must support the scenario where all the instances fail over to the
same node, you need to make sure that the sum of all the 'max server memory'
values is less than the total physical memory minus whatever amount you leave
to the OS and other processes.

You may want to consider just running a single instance to host all 20
databases. We regularly support 20 or more databases on a single instance,
and generally prefer not to go to multiple instances unless we have a
specific reason (e.g. collation conflict). However, running two instances on
a two-node cluster is a pretty common configuration.

Linchi

[quoted text, click to view]