Wrong title for this.
Should be nmon memory output question.


I've ran this by my Linux admins, but can't seem to get a reasonable answer.
I have two systems supposedly having relatively the same memory configurations.

What I'm wondering about is the Cached= and Active= values for each and why they may be so different? Along with the other values underneath. If this could be a difference in configuration or something else.


System1
Memory Stats --------------------------------------------
RAM High Low Swap
Total MB 64425.1 -0.0 -0.0 8192.0
Free MB 54826.1 -0.0 -0.0 8073.7
Free Percent 85.1% 100.0% 100.0% 98.6%
MB MB MB
Cached= 7040.1 Active= 6224.6
Buffers= 425.8 Swapcached= 4.5 Inactive = 2140.5
Dirty = 4.4 Writeback = 0.0 Mapped = 115.9
Slab = 777.5 Commit_AS = 2852.7 PageTables= 18.5
----------------------------------------------------------


System2
Memory Stats --------------------------------------------
RAM High Low Swap
Total MB 64423.5 -0.0 -0.0 8192.0
Free MB 3862.4 -0.0 -0.0 7233.0
Free Percent 6.0% 100.0% 100.0% 88.3%
MB MB MB
Cached= 54948.1 Active= 19277.2
Buffers= 599.8 Swapcached= 69.5 Inactive = 38364.4
Dirty = 16.9 Writeback = 0.1 Mapped = 156.0
Slab = 2380.2 Commit_AS = 6982.8 PageTables= 72.9
----------------------------------------------------------

Your question would be readable if you had used code tags to format nmon output. See my signature for how to do it.

What I see is one system with a small filesystem buffer cache and 85% free memory, and another system with a large filesystem buffer cache and just 6% free memory. If a process needs a lot of memory in the second system, the buffer cache needs to shrink first.

I am not sure what the meaning of "active" is, but my guess is that the non-active pages can easily be removed from the buffer cache, without any file IO.

So the question is how you arrived at this situation. Obviuosly, recently a lot of file I/O took place on system 2, or is still taking place, whereas system 1 is rather quiet.

Memory configuration may be similar, but usage will almost never be on disparate systems.
Cache usage is usually not an issue - by design memory management will attempt to use otherwise idle memory to improve (disk) I/O response times. Slab usage and page table growth on the other hand should really get your attention. Somebody is misbehaving - I'd guess lots (and I mean *lots*) of probably small memory requests. Run these - and use [code] tags as advised when posting (also go back and edit your initial post)Edit: forgot to ask: what distro, what kernel version ?.