memory utilization

[Mohammad Sanaullah]

hi all guru's

Here i am facing a problem on HP Tru64 Server where memory urtilization is more than the 99% and swap utilization is only 6%, i am attaching the vmstat -p output and swapon output.

Please suggest me where my all memories are utilized. also how to make it normal.


Total Physical Memory = 8192.00 M
= 1048576 pages

Physical Memory Clusters:

start_pfn end_pfn type size_pages / size_bytes
0 885 pal 885 / 6.91M
885 1048567 os 1047682 / 8185.02M
1048567 1048576 pal 9 / 72.00k

Physical Memory Use:

start_pfn end_pfn type size_pages/size_bytes

885 1057 scavenge 172 / 1.34M
1057 2216 text 1159 / 9.05M
2216 2480 data 264 / 2.06M
2480 2979 bss 499 / 3.90M
2979 3270 kdebug 291 / 2.27M
3270 3277 cfgmgmt 7 / 56.00k
3277 3279 locks 2 / 16.00k
3279 3293 pmap 14 / 112.00k
3293 6073 unixtable 2780 / 21.72M
6073 6265 logs 192 / 1.50M
6265 30638 vmtables 24373 / 190.41M
30638 1048567 managed 1017929 / 7952.57M
=====================
Total Physical Memory Use:1047682 / 8185.02M

Managed Pages Break Down:

free pages = 4816
active pages = 130373
inactive pages = 100738
wired pages = 113329
ubc pages = 668845
==================
Total = 1018101

WIRED Pages Break Down:

vm wired pages = 11841
ubc wired pages = 6
meta data pages = 31358
malloc pages = 60076
contig pages = 3769
user ptepages = 4854
kernel ptepages = 1051
free ptepages = 10
==================
Total = 112965

Swap partition /dev/disk/dsk15c (default swap):
Allocated space: 2222817 pages (16.96GB)
In-use space: 149081 pages ( 6%)
Free space: 2073736 pages ( 93%)


Total swap allocation:
Allocated space: 2222817 pages (16.96GB)
Reserved space: 361374 pages ( 16%)
In-use space: 149081 pages ( 6%)
Available space: 1861443 pages ( 83%)

[Ivan Ferreira]

As far I can see, the system is using your memory correctly because is not swapping, and most of the memory is used by the UBC (cache for file system objects). The UBC memory is reused if application need it.

You can tune the memory % used by UBC tuning the ubc_maxpercent kernel paramenter.

[Vladimir Fabecic]

Ivan may be right. It looks OK to me also.
Can you post output of:
# sysconfig -q vm

[jim owens_1]

Are you only concerned about memory use because it does not match what some other operating system reports?

If you have a real performance issue, you should describe that issue and the data you have collected. It is almost never as simple as VM use.

Without specific goals and tests that can be repeated to determine the results, changing tuning from the system defaults will only make problems worse.

[Mohammad Sanaullah]

Thanks for responsing to ma question dear all my problem is that on the production server i have 8GB of RAM there is a application running on that server and their vendor says that application is ok network team also says same, i see every thing is ok OS side but i was unable to explain why it shows 99.9% memory is utilised and during vmstat -P output it shows that almost all the RAM is utilised or in use.. but the scene is that i checked the earlier report when there was no application problem then too my server memory is showing 99.9% memory utilised.

then how can i proof that this is not an OS related problem and everything is ok from my side. or is there any problem in my server end???? please suggest!!!

[Rob Leadbeater]

Hi,

Can you explain exactly what the problem is...?

What symptoms are you seeing with the application, other than all of the memory being used ?

Cheers,

Rob

[Vladimir Fabecic]

As said before, your memory is probably used by UBC.
So all unused memory (unused by applications) is used for caching.
This is normal. You are NOT having problem with server.

[Ivan Ferreira]

Please also keep asigning points!

Have a nice day.

[Mohammad Sanaullah]

Dear Rob Leadbeater,

My server is having messaging with one of VAX server and there is no problem at all, but some time it got slow messaging response and when i checked all the CPU, SWAP, MEMORY and SWAP,, there is no difference in report from normal time.
network report also have no problem. but the application team is pointing that why system is utilising the 99.9% Memory which is same as that of normal time to.

Now what should be my course of action on this point.
how can i prove that memory utilisation that is 99.9% have no bad impact on applications running.

that all!!!! hope u all got my point...

Suggest!!>???

[jim owens_1]

As I understand it, there are times when your users complain that response is slower than at other times. This complaint is common when users compare it to what they get as "best response" and expect that is what they should always see.

There are no easy ways to determine what is happening and if the response times match what can be expected from the system configuration.

The only technique I know is to use a tool like "collect" to monitor the system across a wide interval as the load and response changes. Note that any monitor tool also affects the system and runs the risk that the performance will degrade to an unacceptible state by running the tool.

Collecting the data is the easiest part of the problem. Analyzing it is the hard part and requires expert knowledge. I would get someone with more experience to handle that.

[Ivan Ferreira]

>>> How can i prove that memory utilisation that is 99.9% have no bad impact on applications running.

Send to them the "Tru64 UNIX - System Configuration and Tuning" manual, available at the HP Tru64 documentation web page, and ask them to read the sections:

1.5.1.1 Using the UBC
4.4.1.3 Modifying the Percentage of Physical Memory the UBC is Using
11.1.3 Tuning the UBC
12â 1 UBC Memory Allocation

If they are unhappy with UBC memory utilization, change the kernel value menthioned before.

[Mohammad Sanaullah]

Dear All once again can any one jsut tell me what free pages means in the above output of vmstat -P does the UBC pages are also taken as free pages as system frees them whenever applications required the...

if Total Physical Memory=8192.00 M=1048576 pages does that means 1 page equal the 0.007815 MB, if that then my free memory that is
free pages = 4816=4816*0.007815 MB=37MB
ubc pages = 668845 means 5227MB
can i assume that total free memory is 5254MB and rest are Memory Utilised????

Please clear my doubt...

All will get good points......

[Mohammad Sanaullah]

One more thing how to get the page size of the sytem.. i think i have page size of 8k as 2222817 pages allocated shows 16.96 GB that comes only if 2222817*8K = 16.96GB.. then above calculations have to be changed in that respect.

please any one can tell how to know the page size?????

[Ivan Ferreira]

From the same manual menthioned in my previous post:

1.6 Memory Resources

The total amount of physical memory is determined by the capacity of the memory boards installed in your system. The virtual memory (vm) subsystem tracks and manages this memory in 8-KB portions called pages, distributing them among the following areas:

- Static wired memory
- Dynamically wired memory
- Physical memory for processes and data caching

[Mohammad Sanaullah]

Thanks for ur reply??? but dear!!!1

As per your statement one page size is of 8k but some where it is also mentioned 4K is that system specific??? if yes then how i will get what paging size my system is using i am having tru64 v5.1. if there is any command or method to know that will solve my problem.....

thank again in advance

[jim owens_1 · Accepted solution]

Tru64 base memory pages are always 8192 bytes.

The term "page" is not exclusive to VM and may refer to something else.

HP-UX uses 4096 byte memory pages as do other platforms so you will see a lot of 4K page discussions.

[Mohammad Sanaullah]

if thats the fact!!! i can handle situation now better with the customer but it will be better if any one can suggest me the commands to know the same????


thanks

[Ivan Ferreira]

From the output of vmstat

vmstat 5
Virtual Memory Statistics: (pagesize = 8192)


See the the header. No doubts there.

[jim owens_1]

Going back to your vmstat, the order of memory availability is:

free pages = 4816... about 39MB ready for immediate use.

inactive pages = 100738... about 825MB not referenced now and reusable after VM cleanup.

ubc pages = 668845... about 5.4GB that will be reallocated as needed. Note that ubc is where applications have their running code and data that lives on disk unless this is something like an oracle database applicattion.