Windows page file on multiple hard drives

windows hard-drive pagefile

16,429

Solution 1

The most relevant information I could find was an article on optimal configuration of the page file under Windows XP.

The article states that Windows will use the page file located on the volume with the least activity. This means that there is no pattern defined.

So, it is arguable that the most likely way that Windows handles multiple page files is by maintaining a table of where each memory page is located. Where a page winds up depends largely on which volume was least active when it got paged out.

Solution 2

In Windows 8.1 for me it totally ignores the second drive and only swap onto the first one.

Not really useful. That's why I got here since I was googling on it.

16,429

Keltari

01010111 01101111 01110111 00100001 00100000 00100000 01011001 01101111 01110101 00100000 01100100 01100101 01100011 01101111 01100100 01100101 01100100 00100000 01110100 01101000 01101001 01110011 00100000 01100010 01101001 01101110 01100001 01110010 01111001 00100000 01110011 01110100 01110010 01101001 01101110 01100111 00101110 00100000 00100000 01011001 01101111 01110101 00100000 01101000 01100001 01110110 01100101 00100000 01010111 01000001 01011001 00100000 01110100 01101111 00100000 01101101 01110101 01100011 01101000 00100000 01110100 01101001 01101101 01100101 00100000 01101111 01101110 00100000 01111001 01101111 01110101 01110010 00100000 01101000 01100001 01101110 01100100 01110011 00101110

Updated on September 18, 2022

Comments

Keltari almost 2 years

I cant find any documentation or articles online online for this...

how does Windows handle page files on multiple hard drives? Does windows treat them as a concatenation, writing to one till its full then moving to the next? Or does it treat them like a stripe, writing to each one incrementally? Does it write to the first one free? Or perhaps some other method?
Keltari over 12 years

interesting article. However does this still hold true?
Andrew Lambert over 12 years

There's no guarantee that this theory holds true at all, even for XP. It is based entirely on supposition.
surfasb over 12 years

It's doubtful they've changed their handling of multiple page files. I'd bet it still holds true. The memory manager isn't exactly something you make wholesale changes to every version.
Andrew Lambert over 12 years

@surfasb True, but as I said this is only speculation to begin with.
surfasb over 12 years

@Amazed: Plus I see little reason to change it. With the addition of Superfetch, it still makes more sense that the pagefile on the drive with the least activity will see the most action.
Jamie Hanrahan almost 10 years

This isn't supposition. It's straight out of Windows Internals. The "table of where each memory page is located" is simply the page tables.
Jamie Hanrahan almost 10 years

How are you measuring this? I'm not doubting you, just wondering what tool you're using. Perfmon is one way to do it (Pagefile object, % usage counter).
Jamie Hanrahan almost 10 years

Oh, I should mention: The choice is made on every pagefile write (each time the modified list is cleared and moved to the standby list). There is no notion of assigning a process to a pagefile, and a pagefile write can contain pages from many processes. So a process can easily have some pages in one pagefile and other pages in a different pagefile. The page table entry has a four-bit field that indicates which pagefile that page was written to, and the usual PFN field is re-used as the location-within-pagefile. Again, Windows Internals is the source for this.