System Performance

[Ashish Nayyar]

There is system performance issue during peak hours of transaction memory utilisation goes upto 90%

System details is
HP 9000 N4000
A5866A N-Class 550 MHz PA8600 CPU 1.5MB cache
A4923A 1024MB High Density SyncDRAM Memory Mod

Total physical memory in system is 20GB

the current config has Three Virtual memory partitions of 3GB, 4GB and 20GB size in a round robin fashion.

as we are assuming about 20% yearly growth in transaction volume .

What should be the SWAP area that need to be allocate to improve performance.

[Patrick Wallek]

What is your swap area now? Or is that what you mean by your three virtual memory partitions?

The only way swap will improve performance is if you don't already have enough swap area to allow full use of your physical RAM.

If you have that covered, then anything else is extra.

What does your 'swapinfo -tam' output look like? Are you actually paging-out? Have you checked 'vmstat,' specifically the 'po' column, to see what it looks like?

If you are not paging out, and your RAM usage is NOT 100%, then there really isn't anything you can do to improve performance.

You could reduce memory usage by not running so many processes, but I doubt that is a valid solution.

[Ashish Nayyar]

here is the SWAP output

Swap Info
SWAP Information
LVM Boot Information
Boot Definitions for Volume Group /dev/vg00:
Physical Volumes belonging in Root Volume Group:
/dev/dsk/c2t6d0 (0/0/2/0.6.0) -- Boot Disk
/dev/dsk/c1t0d0 (0/8/0/0.0.0)
/dev/dsk/c3t6d0 (0/0/2/1.6.0) -- Boot Disk
Boot: lvol1 on: /dev/dsk/c2t6d0
/dev/dsk/c3t6d0
Root: lvol3 on: /dev/dsk/c2t6d0
/dev/dsk/c3t6d0
Swap: swap1 on: /dev/dsk/c2t6d0
/dev/dsk/c1t0d0
Dump: swap1 on: /dev/dsk/c2t6d0, 0
Dump: lvol2 on: /dev/dsk/c2t6d0, 1
Dump: dump1 on: /dev/dsk/c2t6d0, 2

WARNING: (BEA-co3): Dump space is underconfigured for a FULL crashdump.
WARNING: (BEA-co3): Dump space is adequate for the CRASHCONF configuration.
WARNING: (BEA-c03): Insufficient file space to hold savecore!

DUMP DISKS
Dump: swap1 on: /dev/dsk/c2t6d0, 0
Dump: lvol2 on: /dev/dsk/c2t6d0, 1
Dump: dump1 on: /dev/dsk/c2t6d0, 2

DUMP DATA
SAVECORE : Enabled
SAVECORE DIR: /var/adm/crash
SAVECORE COMPRESSION: Available
COMPRESSION OPTION: N/A
COMPRESSION FACTOR: 50 %
COMPRESSED MEMORY TO DUMP: 10251 MB
SAVECORE CAPACITY: 6197 MB


CRASHCONF : Enabled (referencing /etc/fstab)
Replace Option : CrashConf values REPLACE existing Kernel Definitions.
Included Pages :
Excluded Pages :
FSTAB File Entries :
CRASHCONF Configured MEMORY TO DUMP: 4233 MB

REAL MEMORY: 5242880 KB
MEMORY: 20480 MB
KERNEL SIZE: 20 MB
KERNEL SAFETY FACTOR: 2 MB
TOTAL MEMORY TO DUMP: 20502 MB
DUMP DISK CAPACITY: 7816 MB

SWAP DATA
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 2048 0 2048 0% 0 - 1 /dev/vg00/swap1
dev 68 0 68 0% 0 - 1 /dev/vg00/lvol2
dev 6000 0 6000 0% 0 - 0 /dev/vgswap/swap
dev 17000 0 17000 0% 0 - 0 /dev/vgswap/swap2
dev 5000 0 5000 0% 0 - 0 /dev/vgswap/swap1
memory 15234 3824 11410 25%
total 45350 3824 41526 8% - 0 -

MEMORY: 20480 MB
LOCKABLE RAM: 20160 MB
UN-LOCKABLE RAM: 320 MB
TOTAL SWAP DEVICE SPACE AVAILABLE: 30116 MB
TOTAL SWAP DEVICE SPACE NEEDED: 5120 MB

PRIMARY SWAP DEVICE: swap1
DC_BEGIN SWAPINFO: swapinfo -mdfMt
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 2048 0 2048 0% 0 - 1 /dev/vg00/swap1
dev 68 0 68 0% 0 - 1 /dev/vg00/lvol2
dev 6000 0 6000 0% 0 - 0 /dev/vgswap/swap
dev 17000 0 17000 0% 0 - 0 /dev/vgswap/swap2
dev 5000 0 5000 0% 0 - 0 /dev/vgswap/swap1
memory 15234 3824 11410 25%
total 45350 3824 41526 8% - 0 -
DC_END SWAPINFO: swapinfo -mdfMt

[Patrick Wallek]

It appears that you have approx. 30 GB of device swap configured and none of it is in use, at least at the time you ran the swapinfo command. That amount of swap definitely allows you to fully utilize all of your physical RAM as well.

At this point there is nothing you can do. You will not be able to improve performance by adding more swap since you are not using what you have.

You said initially that there is a system performance issue during peak hours. What is the issue? Are your users complaining about performance? Or is it just that your memory usage goes up to 90%?

In my opinion 90% memory usage is good. You want to use what you paid for, right?

If your users are complaining, you really need to look at other aspects of the system as well.

[Ashish Nayyar]

during peak the system respose gets very
slow .


is there anyways we can enhance system performance.

what default HP ux services should be stopped to improve system

[Ashish Nayyar]

Hi

as u said At this point there is nothing much can be done .It will not be able to improve performance by even after adding more swap since you are not using what you have

what best can be done to use what (SWAP)has been configured.

can some configuration tuning need to be done.

[sreekanthtm]

Hi,
Please understand the swap concept first. If your system is using swap space, it means there is a deficiency in physical memory and you need to upgrade it. Here in your case the system has enough memory and swap is not at all used. Its really good. The performance issue can be due to some other reason... What about the CPU utilization and Buffer cache (look at the sar o/p). And also n/w part also....

Rgds
Sreekanth