Yesterday I ended up running an impromptu browser test, comparing memory utilization. See what Twitter can do to your time!! The browsers themselves are really irrelevant. What you might find useful is the little PowerShell code I put together to periodically check and compare the browser processes.<\/p>\n
I decided to watch the WorkingSet property as a calculation of memory utilization. The value is in bytes. For example, here's the current workingset size for my PowerShell session.<\/p>\n
[cc lang=\"DOS\"]
\nPS S:\\> (get-process -id $pid).workingset
\n236036096
\n[\/cc]<\/p>\n
I can easily convert this to MB.<\/p>\n
[cc lang=\"DOS\"]
\nPS S:\\> (get-process -id $pid).workingset\/1mb
\n225.390625
\n[\/cc]<\/p>\n
I can do the same for any process. The browsers I tested were Firefox 4, Internet Explorer 9 and the latest version of Chrome. All had the same 4 sites opened and were pretty much left untouched. To calculate the memory usage I can do the same for each process. But there is one potential challenge. Firefox runs everything in a single process so I could use the same technique. However, the other two browsers run multiple instances of the same process. So what I have to do is add up the working set property for all the processes and then divide by 1MB. I can do this in one line.<\/p>\n
[cc lang=\"DOS\"]
\nPS S:\\> (get-process chrome | measure-object ws -sum).sum\/1mb
\n369.71875
\n[\/cc]<\/p>\n
All of the Chrome processes are piped to Measure-Object which calculates the sum on the ws property (which is an alias for Workingset). That returns a measurement object which I \"get\" by wrapping the expression in parentheses. I can then get the sum property value and divide by 1MB. That's the core comparison. But I had to do that for each browser process. So I put together a short script to make this more meaningful and easier to run.<\/p>\n
[cc lang=\"PowerShell\"]
\n#Requires -version 2.0<\/p>\n
#array of process names
\n$Procs=\"iexplore\",\"chrome\",\"firefox\"<\/p>\n
$Procs | foreach {
\n #get the earliest start time
\n $started=Get-Process $_ | sort StartTime | Select -first 1
\n #create a custom object
\n New-Object PSObject -Property @{
\n Name=$_
\n WS=(get-process $_ | measure-object ws -sum).sum\/1mb
\n Report=Get-Date
\n Started=$started.StartTime
\n Elapsed=(Get-Date)-($Started.StartTime)
\n }
\n} #foreach
\n[\/cc]<\/p>\n
The script takes an array of process names and pipes them to ForEach-Object. Each process gets measured. I also get when the process started. For browsers with multiple processes, I get the process with the earliest starttime. With this I can calculate a run time. These values are assembled into a new object and written to the pipeline. Here's the end result.<\/p>\n
[cc lang=\"DOS\"]
\nPS S:\\> C:\\work\\browsertest.ps1 | sort ws<\/p>\n
Name : firefox
\nStarted : 5\/10\/2011 12:20:13 PM
\nWS : 252.93359375
\nElapsed : 20:14:02.8323717
\nReport : 5\/11\/2011 8:34:16 AM<\/p>\n
Name : iexplore
\nStarted : 5\/10\/2011 6:09:57 PM
\nWS : 314.6015625
\nElapsed : 14:24:19.1541740
\nReport : 5\/11\/2011 8:34:16 AM<\/p>\n
Name : chrome
\nStarted : 5\/10\/2011 12:19:43 PM
\nWS : 369.12109375
\nElapsed : 20:14:33.1991086
\nReport : 5\/11\/2011 8:34:16 AM
\n[\/cc]<\/p>\n
By creating objects I now have opportunity. For example, I could periodically run this command and export the results to a CSV file which I could then open in Microsoft Excel and analyze or chart. You could easily turn this into a function specifying any process names and even remote computer names. Although don't forget to add some error handling for non-existent processes.<\/p>\n