With vSphere 8.0, a couple of new performance counters have been implemented in vCenter Server. When you migrate a Virtual Machine to another Host, the hostd service running on the ESXi Host keeps track of performance statistics from the vMotion process. These statistics can now be viewed in the vSphere Clients performance pane.
Using synthetic workloads to test drive PernixData FVP might result into odd findings. The most meaningful approach to test FVP is to deploy the software to production in monitor mode, let Architect do its magic and enable acceleration after checking the recommendations after a couple of days. Despite it is possible to deploy FVP, test drive, and remove it, without any downtime to virtual machines, this approach might not fit to all environments.
If you have separate DEV/QA environments with sophisticated load generators, the solution is obviously. If you don't have that, there is another option by record production I/O traces and replay them in a FVP accelerated test platform.
I've read this excellent article by Cormac Hogan explaining why and when a Virtual Machine receives a "stun". This post is a follow-up explaining how you can determine the duration of a Virtual Machine stun on the most common vSphere functions:
- Create a Snapshot
- Delete a Snapshot
When analysing performance metrics in the vSphere Web Client (Monitor > Performance) or in the vSphere Client (Performance Tab) you might have seen the following messages instead of performance graphs:
"No data available"
"Data is not collected for the current statistics level. Increase the statistics level to view the graph."
Interval Duration determines the frequency at which statistics are stores:
- Realtime (20 seconds), save for 1 hour - not configurable, all metrics available
- 5 minutes, save for X days
- 30 minutes, save for X weeks
- 2 hours, save for X month
- 1 day, save for X month
Statistic Level determines the amount of data gathered and which counters are available for displayed. The default Level 1 stores the fewest metrics, Level 4 stores all metrics supported by the vCenter Server.
With vSphere 6.0, vCenter Server supports 544 metrics but as there are only 4 statistic levels it is not clear what metrics are included in each level. This post helps to understand what metrics are included in each level, and how you can add single metrics to lower levels. This might be helpful if you need single metrics from level 2, but do not want to activate all level 2 metrics.
Please note that changing the collection level beyond level 1 or adding a large number of data counters to collection level 1, might result in a significant reduced performance.
This is a list of all available performance metrics that are available in vSphere vCenter Server 6.0. Performance counters can be views for Virtual Machines, Hosts, Clusters, Resource Pools and other objects by opening Monitor > Performance in the vSphere Web Client.
These performance counters can also be used for performance analysis with esxcfg-perf.pl, or PowerCLI.
I haven't seen this in any announcement, but there is a new performance metric in vSphere 6.0 called CPU Readiness which is according to the description:
Percent of time the virtual machine is unable to run because it is contending for access to the physical CPU(s).