So you may ask how can we determine the total required data center memory?

Well, there are several methods to determine the total memory capacity requirement:

• Use the information gathered during the current state analysis to determine the current memory capacity requirements
• Use application vendor documentation to estimate memory capacity requirements
• Actual data center usage analysis is typically more accurate

Do not plan on fully utilizing host memory resources. You will want to have enough memory so you can remove one host from any cluster and still run all the VMs. This will allow you to perform maintenance on a host or have a host fail without the need to power off VMs. Also, leave some headroom for such things as:

• Increase in short-term utilization as part of the normal business cycle
• Failover from other data centers during disaster recovery
• VMkernel overhead
• VM overhead
• Future growth

In this chapter, we will cover the tasks related to memory performance design. You will learn the following aspects of memory performance design:

• Virtual memory reclamation techniques
• Monitoring a host-swapping activity
• Monitoring a host-ballooning activity
• Keeping memory free for VMkernel
• Key memory performance counters to monitor
• What counters not to use
• Identifying when memory is the problem
• Analyzing host and VM memory
• Memory performance best practices