Monitoring - jasper-zanjani/Windows-Server GitHub Wiki

Performance Monitor is a program that allows realtime monitoring of hundreds of different system performance statistics, called performance counters. Counters can be viewed in several ways, including line graph, histogram bar graph, and report views. Every counter added to a graph is associated with a computer, a performance object (hardware or software component to be monitored), a performance counter (statistic), and an instance.

A data collector set captures counter statistics for later review. A single data collector set can gather performance counter data from multiple VMs. Event trace data cannot be combined with performance data in the same data collector set. Expiration dates can be set for data collector sets, but if actively collecting data the expiration date will not stop collection.

A performance alert is a type of data collector set that can track system performance and log events in the application event log.

Alerts can be triggered when a performance counter value exceeds a certain threshold. Only members of the local groups Administrators and Performance Log Users can create alerts, but the Log on as a batch user right must be granted to members of Performance Log Users.

A hard fault occurs when data is swapped between memory and disk.

Performance counters

Counter	Acceptable values
Processor: % Processor Time	<85%
Processor: Interrupts/Sec	cf. baseline
System: Processor Queue Length	<2
Server Work Queues: Queue Length	<4
Memory: Page Faults/Sec	<5
Memory: Pages/Sec	<20
Memory: Available MBytes	>5% of physical memory
Memory: Committed Bytes	< physical memory
Memory: Pool Non-Paged Bytes	Stable
PhysicalDisk: Disk Bytes/Sec	cf. baseline
PhysicalDisk: Avg. Disk Bytes/Transfer	cf. baseline
PhysicalDisk: Current Disk Queue Length	<2 per spindle
PhysicalDisk: % Disk Time	<90%
LogicalDisk: % Free Space	>20%
Network Interface: Bytes Total/Sec	cf. baseline
Network Interface: Output Queue Length	<2
Server: Bytes Total/Sec	50% of total bandwidth