Understanding Hard & Soft Page Faults with ControlUp

A user asked about hard and soft page faults and ControlUp’s reporting capabilities regarding these faults. Hard page faults generally have a larger negative impact on performance, however with more people using solid state drives this impact is not as drastic. ControlUp is working to include this measurement of hard page faults in its reporting. In terms of acceptable page fault expectations, it varies depending on underlying hardware and disk speeds, and looking at the combination of hard page faults and excessive queue lengths can help to determine if there is an issue.

Read the entire ‘Understanding Hard & Soft Page Faults with ControlUp’ thread below:

Hello, a stupid question maybe, I have provided my security team with metrics about Sentinel, where I prove that it’s consuming a lot of CPU and RAM, and doing tons of page faulting.

I am asked if it’s hard or soft page fautling… Any help ?

Thanks in advance, cheers, JCM

Thats new to me but i found these details

• Hard Page Fault – The application memory-page has been paged to disk, and now it has to be loaded from disk. This is a blocking operation and will hurt application performance.

• Soft Page Fault – The application memory-page resides in the standby-list, and can be quickly loaded back into the application working set.

Soft Page faults also happens when the application allocates new memory-pages, by taking them from the free-list / standby-list.

thanks @member, then what is ControlUp actually reporting ?

"Total Page Faults / sec consumed by this executable on all machines" [1]

I guess ControlUp is using the Process\Page Faults/s Total performance counter…

But its just a guess maybe @member can verify 🙂

[1] https://support.controlup.com/docs/applications-view

So… bit late here. But on the Hard/Soft page fault thing. Windows used to have separate Hard and Soft page fault metrics, but at some point the Hard Page faults were dropped.

Regarding this: ‘Hard Page Fault – The application memory-page has been paged to disk, and now it has to be loaded from disk. This is a blocking operation and will hurt application performance.’ –> yes, Hard Page Faults do have a bigger impact than soft page faults. But the impact has changed a lot now nearly everyone is on solid state instead of spinning disks. They still have a bigger impact but not nearly as bad as it used to be. Hard Paging can even be a good thing, it is the process of freeing up physical memory by paging it out (and reading it back to memory when it is required).

Tl;dr: high hard page faults still not good, not half as bad as it used to be though.

@member’s assumption about which counter we use is correct. But. You heard it here first folks, we are working on displaying Hard Page Faults as a new metric. No promises on ETA though 🙂Hello, following up on this, is there a good article on what is the right or acceptable amount of page faults for a process in a VDI with x amount of memory, x CPUs, etc…? Just asking…

It is really tricky. A lot depends on the underlying hardware. If you have a lot of page faults but your disk is the latest NvmE and it serves up those pages in a split second, it may not be a problem at all. Old laptop 5400 RPM disk, that changes things.

What I would look at is a combination of hard page faults and excessive queue lengths. Aaaaand what que lengths are excessive is also subjective… Look for combinations that do not match the pattern on machines that perform well.

Sidenote: did you find a good calculation for getting those hard page faults? I’ve been down a rabbit hole a bit on this one, hard to find definitive answers.

Thanks @member, I’ll cross check with machines that perform ok.

Someone must have an answer to your last question…

There are answers, but not easy ones 🙂

Continue reading and comment on the thread ‘Understanding Hard & Soft Page Faults with ControlUp’.  Not a member? Join Here!