126 reputation
5
bio website stackoverflow.com/users/37941/…
location London, United Kingdom
age 36
visits member for 4 years, 2 months
seen Nov 4 at 15:16

About Me? Are you kidding? I'm only filling this in for the Autobiographer badge :)


Jun
16
awarded  Popular Question
Jun
26
comment How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
I can't I'm afraid, but I appreciate the offer. You can try your own example - your VMMap screenshot says Private Bytes=460MB (second bar graph) whereas procexp reports 200MB - so you are seeing the same problem already (to a lesser degree). I bet when you kill the process that the impact on commit charge is ~460 MB.
Jun
25
comment How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
I'm not trying to start learning about windows memory management - I already know some, and I'm trying to understand why SQL server violates everything that I already thought I knew! I've deliberately picked an example that doesn't behave as expected in the hope that the normal tools (eg procexp) can help identify issues. Right now if I have the problem that a server has a huge commit charge but very little committed private bytes across processes, to identify which process is responsible I have to run VMMap against every single process because procexp 'lies' :-(
Jun
25
comment How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
FWIW, I don't believe you can write stuff to reserved, uncomitted memory anyway (for "caching database data and query cache plans and junk") - you have to commit the reserved memory before you can actually use it.
Jun
25
revised How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
Emphasising 'committed', adding windows tags
Jun
25
comment How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
@Ryan - this still doesn't address the Q! :-( You're looking at reserved Private Data, I'm fully aware that reserved memory can be massive and the only meaningful thing to compare it to is the Virtual Size in procexp. I'm specifically highlighting the committed Private Data (440MB in your screenshot) - not reserved!! (i.e. nothing to do with SQl server "reserving all the memory it can") vs the 200MB Private Bytes you see in procexp. (In my example, the commmitted Private Data is 1.6GB).
Jun
24
revised How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
reworded title
Jun
24
revised How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
corrected typo
Jun
24
awarded  Editor
Jun
24
revised How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
added 993 characters in body
Jun
24
comment How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
@Mark - RamMap doesn't work on Server 2003 :-( I have been looking at VMMap though, I'll update my question with findings...
Jun
24
awarded  Student
Jun
24
asked How can the Private Bytes of a process be significantly less than its effect on the system commit charge?
Oct
10
awarded  Autobiographer
Oct
10
awarded  Supporter