This document contains tips for allocating paging space on the system. The information contained in this document is valid for 4.3.X.
To obtain more information, see the AIX Performance Management Guide, available online at:
http://www.rs6000.ibm.com/doc_link/en_US/a_doc_lib/aixbman/prftungd/toc.htm
The product documentation library is also available:
http://www.rs6000.ibm.com/resource/aix_resource/Pubs/index.html
The following can be found at:
http://9.53.35.177:80/techlib/manuals/adoclib/aixbman/baseadmn/toc.htm AIX Version 4.3 System Management Guide: Operating System and Devices Chapter 8.
Paging Space and Virtual Memory
The amount of paging space required depends on the types of activities performed on the system. If paging space runs low, processes may be lost, and if paging space runs out, the system may panic. When a paging-space low condition is detected, additional paging space should be activated.
The system monitors the number of free paging space blocks and detects when a paging space shortage exists. When the number of free paging space blocks falls below a threshold known as the paging space warning level, the system sends the SIGDANGER signal to all processes except the kprocs process. If the shortage continues, free paging space blocks can fall below a second threshold known as the paging space kill level. In this event, the SIGKILL signal is sent to processes that are major users of paging space and do not have a signal handler for the SIGDANGER signal. (The default action for the SIGDANGER signal is to ignore the signal.) The system continues sending SIGKILL signals until the number of free paging space blocks is above the paging space kill level.
You can ensure the existence of sufficient paging space for processes that dynamically allocate memory by monitoring the paging space levels with the psdanger subroutine or by using special allocation routines. The disclaim subroutine can be used to prevent processes from ending when the paging space kill level is reached. To do this, define a signal handler for the SIGDANGER signal and release memory and paging space resources allocated in the processes data and stack areas and in shared memory segments.
For more information on persistent and working segments, request The AIX Virtual Memory Manager (VMM) document.
Paging space requirements are unique for each system, depending on the applications that are running, the number of active users, and other factors. With the appearance of large amounts of RAM and database applications, previous paging space rules of thumb have become invalid.
Option 1
Systems with large amounts of memory typically do not need large amounts of paging space. In a persistent storage environment, where the system hosts a few small programs and a large amount of data, the system may need less than one times (1X) its RAM size for paging space. For example, a 100GB database server that runs on a system with 16GB of RAM and uses only 2GB of working storage does not need 16GB, or even 8GB, of paging space. Because the 100GB database is mostly persistent storage and requires little or no paging space, it needs only the amount of paging space that allows all the working storage to be paged out to disk.
Option 2
The 1X RAM rule is suggested for use with less than or equal to 4GB of RAM. However, the paging space will have to be monitored during a period of heavy load to establish whether the paging space size is sufficient. Check the npswarn value of the vmtune command output and compare this value to the %used value of the lsps command output. When the %used value is equivalent to the npswarn value, then SIGDANGER signals are sent to processes. At this point, it is a good idea to increase or add another paging space. The npswarn value is explained in the Tuning paging space thresholds section of this document.
Option 3
For RAM sizes greater than 4GB, such as 16GB, 32GB or even 96GB, memory requirements for applications would have to be researched in order to approximate the recommended paging space sizes. When researching these memory requirements , keep in mind how the paging space will be allocated, that is, deferred or late.
Before creating a new paging space or enlarging an existing paging space, consider the following:
NOTE: If system is paging enough to cause an I/O bottleneck, tuning the location of the paging space is not the answer.
In this case, please consult Chapter 7, "Monitoring and Tuning Memory Use of Performance Management Guide", in the Performance Management Guide at this location:
http://www.rs6000.ibm.com/doc_link/en_US/a_doc_lib/aixbman/prftungd/toc.htm
Allocating more paging space than necessary results in unused paging space that wastes disk space. However, allocating too little paging space can result in one or more of the avoidable symptoms listed below. Use the following guidelines for determining the necessary paging space:
| INIT: Paging space is low |
| ksh: cannot fork no swap space |
| Not enough memory |
| Fork function failed |
| fork () system call failed |
| Unable to fork, too many processes |
| Fork failure - not enough memory available |
| Fork function not allowed. Not enough memory available. |
| Cannot fork: Not enough space |
| SIGKILL |
NOTE: Only extend a paging space as a last option.
Use the following commands to determine if you need to make changes regarding paging space logical volumes:
| iostat | Check the tm_act field for the hdisk containing the paging space for a high percentage relative to the other hdisks |
| vmstat | Assure fr/sr columns of the vmstat page field do not consistently exceed the ratio of 1:4. |
| lsps | Use the -a flag to list all characteristics of all paging spaces. The size is given in megabytes. Use the -s flag to list the summary characteristics of all paging spaces. This information consists of the total paging space in megabytes and the percentage of paging space currently assigned (used). If the -s flag is specified, all other flags are ignored. |
Follow these steps to add a paging space. Please note that the command output shown in this example may differ from the command output on your system.
The following command lists characteristics for all existing paging spaces:
#lsps -a Page Space Physical Volume Volume Group Size %Used Active Auto Type Paging02 hdisk2 rootvg 512MB 3 yes yes lv Paging01 hdisk3 testcase 512MB 3 yes yes lv Paging00 hdisk1 rootvg 512MB 2 yes yes lv Hd6 hdisk0 rootvg 512MB 3 yes yes lv
The following command lists the available physical volumes:
#lspv hdisk0 000336524e264c40 rootvg hdisk1 00033652f9fe5c81 doomvg hdisk2 00302593eb30798f none hdisk3 00033652fa08edca testcase
The following command displays detailed information about the physical volume within a volume group where you plan to assign the paging space.
#lspv hdisk3 PHYSICAL VOLUME: hdisk3 VOLUME GROUP: testcase PV IDENTIFIER: 00033652fa08edca VG IDENTIFIER 0008508436f7d210 PV STATE: active STALE PARTITIONS: 0 ALLOCATABLE: yes PP SIZE: 8 megabyte(s) LOGICAL VOLUMES: 3 TOTAL PPs: 268 (2144 megabytes) VG DESCRIPTORS: 2 FREE PPs: 201 (1608 megabytes) USED PPs: 67 (536 megabytes) FREE DISTRIBUTION: 00..00..00..00..07 USED DISTRIBUTION: 54..54..53..53..47
Do one of the following:
#smitty mkps
OR
#smitty >>System Storage Management >>Logical Volume Manager >>Paging Space >>Add another paging space.
# of logical = Prospective Paging Space Size in MBs Partitions PP SIZE
No reboot is required, and can you can complete this procedure on a production system.
Follow these steps to increase the size of an existing paging space:
smitty chps
For additional information, please Chapter 7, "Monitoring and Tuning Memory Use", in the Performance Management Guide at this location:
http://www.rs6000.ibm.com/doc_link/en_US/a_doc_lib/aixbman/prftungd/toc.htm
If available paging space becomes depleted, the operating system attempts to release resources as follows:
The VMM uses the values of two parameters that specify the thresholds for sending warning or kill signals to processes:
| npswarn | The value of this parameter specifies the paging space warning threshold, below which warning signals are sent to processes. |
| npskill | The value of this parameter specifies the paging space kill threshold, below which kill signals are sent to certain processes. |
These values for the npswarn and npskill parameters are set by means of arguments to the vmtune command:
| Parameter | vmtune flag | Description |
|---|---|---|
| npswarn | -w |
Specifies the number of free paging space pages at which the operating system
begins sending the SIGDANGER signal to processes. If the npswarn threshold
is reached and a process is handling this
signal, the process can choose to
ignore the signal or do some other action, such as
exit or release memory by using the
disclaim() subroutine. The default value
in operating system version 4 is
determined by the following formula:
npswarn = 4*npskill The value of npswarn must be greater than zero and less than the total number of paging space pages on the system. This parameter can be changed by using the vmtune -w command. |
| npskill | -k |
Specifies the number of free paging space pages at which the operating system
begins killing processes. If the npskill threshold is reached, a SIGKILL
signal is sent to the youngest process. Processes that are handling SIGDANGER
or processes that are using the early page-space allocation (paging space is
allocated as soon as memory is requested) are exempt from being killed. The
formula to determine the default value of npskill is as follows:
npskill = number_of_paging_space_pages/128 The npskill value must be greater than zero and less than the total number of paging space pages on the system. This parameter can be changed by using the vmtune -k command. |
Real memory = 16GB
Paging space = 4096MB
Convert paging space to 4KB pages.
4096MB *(1024KB/4KB)=1048576 4KB pages
npskill = number_of_paging_space_pages/128
= 1048576 pages/128
= 8192 pages
npswarn = 4*npskill
= 4*8192
= 32768 pages
Npskill % of paging space = ((1048576 - 8192)/1048576)*100
= 99.2
Npswarn % of paging space = ((1048576 - 32768)/1048576)*100
= 96.8
Example of a 8GB paging space:
Npskill % of paging space = ((2097152 - 16384)/2097152)*100
= 99.2
Npswarn % of paging space = ((2097152 - 65536)/2097152)*100
= 96.8
Example of a 16GB paging space:
Npskill % of paging space = ((4194304 - 32768)/4194304)*100
= 99.2
Npswarn % of paging space =((4194304 - 131072)/4194304)*100
= 96.8
The npswarn and npskill default value percentages for paging spaces of 4GB, 8GB and 16GB all are equivalent. Notice the npswarn default is 96.8%. This percentage translates to the point at which the paging space usage percentage will send SIGDANGER signals to marked processes. Then when the paging space usage percentage reaches 99.2, the SIGKILL signal will sent to the youngest process. To gain more notice time, decrease npswarn percentages to give a more advanced warning.
| Parameter | vmtune flag | Description |
|---|---|---|
| nokillroot | -n | By setting the nokillroot option to 1 with the command vmtune -n 1, processes owned by root will be exempt from being killed when the npskill threshold is reached. This option is only available in operating system version 4.3.3 and 4.3.3.1. |
| nokilluid | -n | By setting the nokilluid option to a nonzero value with the command vmtune -n, user IDs lower than this value will be exempt from being killed because of low page-space conditions. This option is only available in operating system version 4.3.3.2 and later. |
If a process cannot be forked due to a lack of paging-space pages, the scheduler retries the fork five times. After each try, the scheduler delays for a default of 10 clock ticks.
The -f flag of the schedtune command specifies the number of clock ticks to wait before retrying a failed fork() call. For example, if a fork() subroutine call fails because there is not enough space available to create a new process, the system retries the call after waiting the specified number of clock ticks. The default value is 10, and because there is one clock tick every 10 ms, the system retries the fork() call every 100 ms.
If the paging space is only low due to brief, sporadic workload peaks, increasing the retry interval might allow processes to delay long enough to be released. See the following example:
# /usr/samples/kernel/schedtune -f 15
In this way, when the system retries the fork() call, there is a higher chance of success because some processes might have finished their execution and, consequently, released pages from paging space.
[ Doc Ref: 90605219714826 Publish Date: Mar. 27, 2001 4FAX Ref: 1075 ]