One question that came up at the summit was the comparative speed of execution of the new CIM cmdlets vs the old WMI cmdlets. No of us knew the answer because we’d never tried measuring the speed.
I decided to perform some tests.
This first test is accessing the local machine. In both cases the cmdlets are using COM. WMI uses COM and CIM will use COM if a –ComputerName parameter isn’t used.
The results are as follows:
PS> 1..100 |
foreach {Measure-Command -Expression {
1..100 | foreach {Get-WmiObject -Class Win32_ComputerSystem} }
} | Measure-Object -Average TotalMilliseconds
Count : 100
Average : 2008.953978
Sum :
Maximum :
Minimum :
Property : TotalMilliseconds
PS> 1..100 |
foreach {Measure-Command -Expression {
1..100 | foreach {Get-CimInstance -ClassName Win32_ComputerSystem} }
} | Measure-Object -Average TotalMilliseconds
Count : 100
Average : 2078.763174
Sum :
Maximum :
Minimum :
Property : TotalMilliseconds
So for pure COM access the WMI cmdlets are marginally (3.4%) faster.
What if we use the ComputerName parameter?
PS> 1..100 |
foreach {
Measure-Command -Expression {
1..100 | foreach {Get-WmiObject -Class Win32_ComputerSystem -ComputerName $env:COMPUTERNAME } }
} | Measure-Object -Average TotalMilliseconds
Count : 100
Average : 1499.14379
Sum :
Maximum :
Minimum :
Property : TotalMilliseconds
PS> 1..100 |
foreach {
Measure-Command -Expression {
1..100 | foreach {Get-CimInstance -ClassName Win32_ComputerSystem -ComputerName $env:COMPUTERNAME } }
} | Measure-Object -Average TotalMilliseconds
Count : 100
Average : 3892.921851
Sum :
Maximum :
Minimum :
Property : TotalMilliseconds
This one surprised me – the WMI cmdlets are 2.5 times faster. I suspect that is because the CIM cmdlet has to build and then breakdown the WSMAN connection each time.
Next time we’ll look at accessing a remote machine.
