Pages/second vs hard faults/second

Devilsbane

From what I've always been told, these are the same thing. I was working a transcender simulation where I was monitoring the memory and page file, and I chose to monitor the hard faults/seconds every 2 minutes and send an alert when the count was over 20. I was supposed to configure an alert when the count was over 20 for pages/seconds. So I received no points.

Just trying to track down the difference between the two counters.

Thanks

gateway

I think this explains it quite well

What are Hard Faults per Second, and How Many of Them is Too Many?

Devilsbane

Still not answering my question.

Here is what I know (or at least think I know). A page is a 4k chunck of information that resides in memory. Ideally, you would have an infinite amount of RAM and you could always keep it there. Since you don't, when your memory starts to become full, it transfers the pages to HDD. The goal is to only transfer pages that haven't been accessing in awhile since the HDD is about 1,000 times slower than your RAM. When the system needs some information, and it has been paged to the HDD and needs to retrieve it, this is a Hard Fault. Hard faults are generally considered bad, since it is going to take extra time, and can also slow down your disk writing.

The only thing that I can think of is that pages/second measures hard page faults and page faults/second measures both hard faults and soft faults. Is that correct?

Devilsbane

After research I have determined that pages/second refers only to hard faults (information needed is on disk) while faults/second refers to both hard and soft faults (soft faults occur when the information is stored in RAM, just a different area)

MS isn't concerned with Soft Faults because it is still exceptionally fast, thefore the faults/second counter will generally not be used. We are mainly concerned with hard faults, because that slows things down (HDD is slow) and will hurt our disk IOs.

RobertKaucher

Devilsbane wrote: »

A page is a 4k chunck of information that resides in memory.

Trick question: Are you sure this is always true?

Also, have you seen this KB?
RAM, Virtual Memory, Pagefile and all that stuff

I highly recommend the Memory Mgmt chapter in the Windows Internals book. If you do not own a copy of that, you should invest in one. It is more for devs but is an amazing resource for us admin types as well.

Devilsbane

RobertKaucher wrote: »

Trick question: Are you sure this is always true?

Are you referring to the potential to swap an entire virtual address space?

And no, I wasn't sure of this. This is what I've been told and this is what I've seen in practice. I really wasn't too concerned with the number, more concerned with the counter.

I also had never seen that KB before. An interesting read, but I don't really have the time to look at every piece of it. I would love to spend more time on this issue, but at the moment I'm concerned with the 290 tomorrow and don't believe this issue will be on the test.

RobertKaucher

Devilsbane wrote: »

Are you referring to the potential to swap an entire virtual address space?

And no, I wasn't sure of this. This is what I've been told and this is what I've seen in practice. I really wasn't too concerned with the number, more concerned with the counter.

I also had never seen that KB before. An interesting read, but I don't really have the time to look at every piece of it. I would love to spend more time on this issue, but at the moment I'm concerned with the 290 tomorrow and don't believe this issue will be on the test.

Yes, certainly not going to find that level of detail on the exam.

It's a minor side note but the architecture for the, now discontinued, Itanium systems (IA64) used 8 KB memory pages.