Difference between revisions of "Paging"

From Hackepedia
Jump to navigationJump to search
(Added example and pagesize table)
 
Line 1: Line 1:
−
When a [[process]] causes a page fault because of missing data, this data is "paged in".  When a process is forfeiting memory the data is "paged out" to swap, subsequently when this data is needed again it is "paged in" again.  Paging is different from swapping because swapping moves the entire memory of a process to/from swap.  Different architectures have different size of pages (pagesize):
+
When a [[process]] causes a page fault because of missing data, this data is "paged in".  When a process is forfeiting memory the data is "paged out" to swap, subsequently when this data is needed again it is "paged in" again.  Paging is different from swapping because swapping moves the entire memory of a process to/from swap.  Paging is performed in units of that are of ''pagesize'' bytes.  Different architectures have different sizes of pages.
  
−
i386:  4096 bytes
+
{|style="width:50%; height:200px align=center margin: 1em auto 1em auto"" border="1" cellpadding=5
−
 
+
|-
−
amd64: 4096 bytes
+
|'''Architectures''' || '''Pagesize (bytes)'''
−
 
+
|-
−
sparc: 4096 or 8192 bytes
+
|i386 || 4096
−
 
+
|-
−
sparc64: 8192 bytes
+
|amd64|| 4096  
−
 
+
|-
−
vax: 4096 bytes
+
|sparc || 4096 or 8192  
−
 
+
|-
−
alpha: 8192 bytes
+
|sparc64 ||8192  
−
 
+
|-
−
macppc: 4096 bytes
+
|vax || 4096  
−
 
+
|-
−
mac68k: 4096 bytes
+
|alpha ||8192  
 +
|-
 +
|macppc ||4096  
 +
|-
 +
|mac68k || 4096  
 +
|}
  
 
To find the pagesize of your system try typing "pagesize".
 
To find the pagesize of your system try typing "pagesize".
Line 21: Line 26:
 
  $ pagesize  
 
  $ pagesize  
 
  4096
 
  4096
 +
 +
Here is sample output from a Solaris 8 production system that is experiencing mild paging, but zero [[swapping]]:
 +
 +
  [root@soemhost /tmp]# vmstat -S 60
 +
  procs    memory            page            disk          faults      cpu
 +
  r b w  swap  free  si  so pi po fr de sr m1 m1 m1 m2  in  sy  cs us sy id
 +
  0 0 0 1591144 1483816 0  0 574 2  2  0  0  0  0  0  1  843 1729  457  6 13 82
 +
  0 0 0 1164512 1051184 0  0 42  6  6  0  0  0  0  0  0  729 2268  222  3 30 66
 +
  0 0 0 1253864 1060440 0  0 17  3  3  0  0  2  2  2  1  500 1977  217  6 33 60
 +
  0 0 0 1208336 1058672 0  0 50  8  8  0  0  1  1  1  0  681 2947  252  5 35 60

Latest revision as of 13:00, 19 April 2006

When a process causes a page fault because of missing data, this data is "paged in". When a process is forfeiting memory the data is "paged out" to swap, subsequently when this data is needed again it is "paged in" again. Paging is different from swapping because swapping moves the entire memory of a process to/from swap. Paging is performed in units of that are of pagesize bytes. Different architectures have different sizes of pages.

Architectures Pagesize (bytes)
i386 4096
amd64 4096
sparc 4096 or 8192
sparc64 8192
vax 4096
alpha 8192
macppc 4096
mac68k 4096

To find the pagesize of your system try typing "pagesize". 

$ pagesize 
4096

Here is sample output from a Solaris 8 production system that is experiencing mild paging, but zero swapping: 

 [root@soemhost /tmp]# vmstat -S 60
  procs     memory            page            disk          faults      cpu
  r b w   swap  free  si  so pi po fr de sr m1 m1 m1 m2   in   sy   cs us sy id
  0 0 0 1591144 1483816 0  0 574 2  2  0  0  0  0  0  1  843 1729  457  6 13 82
  0 0 0 1164512 1051184 0  0 42  6  6  0  0  0  0  0  0  729 2268  222  3 30 66
  0 0 0 1253864 1060440 0  0 17  3  3  0  0  2  2  2  1  500 1977  217  6 33 60
  0 0 0 1208336 1058672 0  0 50  8  8  0  0  1  1  1  0  681 2947  252  5 35 60
Retrieved from "https://hackepedia.org/?title=Paging&oldid=2711"