Does it make sense to add more cluster nodes to an underutilized server?

Let's say I'm running a cluster with two Weblogic server nodes on two servers (Solaris, 4 CPUs, 64 GB RAM each). The heap size is maxed out on these nodes and so each server runs with 40+ GB of free memory most of the time and each server usually has on average 30% CPU utilization.

Assuming that the application scales horizontally properly, does it make sense to add more ap开发者_StackOverflowplication server nodes on the same hardware to take advantage of underutilized resources? The 2 GB JVM heap limit really limits the maximum number of requests that can be served by one node, right?

The heap size is maxed out on these nodes and so each server runs with 40+ GB of free memory most of the time and each server usually has on average 30% CPU utilization.

Under what load? Maybe the current load is just too low.

(...) does it make sense to add more application server nodes on the same hardware to take advantage of underutilized resources?

Vertical clustering allows indeed to better utilize memory or threading (and gives you protection against software failure).

The 2 GB JVM heap limit really limits the maximum number of requests that can be served by one node, right?

Threads do consume memory but you are actually more limited by the power of your processor(s). See:

• Tuning WebLogic Server (WLS 8.1)
  • Tuning the Default Execute Queue Threads (still an interesting reading, even if you are using later versions and work managers)
• Tuning WebLogic Server
  • Thread Management
  • Performance Considerations When Running Multiple Server Instances on Multi-CPU Machines