-
vmstat 10 10
procs memory page disk faults cpu
r b w swap free re mf pi po fr de sr s0 s1 s1 -- in sy cs us sy id
0 0 0 14216 8504 0 1 1 8 8 0 0 0 1 0 0 218 480 704 3 2 95
0 0 0 872648 7944 0 1 0 20 32 0 5 0 1 0 0 209 977 694 0 0 99
0 0 0 872648 7944 0 0 0 7 7 0 0 3 2 0 0 227 886 694 0 0 100
0 0 0 872648 7952 0 0 0 8 8 0 0 0 1 0 0 209 973 694 0 0 99
0 0 0 872648 7960 1 0 0 9 9 0 0 0 1 0 0 208 860 692 0 0 100
0 0 0 872648 7976 0 0 0 10 10 0 0 1 1 0 0 219 974 695 0 0 99
0 0 0 872648 7984 0 0 0 8 8 0 0 0 1 0 0 207 870 691 0 0 100
0 0 0 872648 7992 0 0 0 7 7 0 0 0 1 0 0 208 998 695 0 1 99
0 0 0 872640 7984 0 0 0 7 7 0 0 1 1 0 0 213 859 692 0 0 100
0 0 0 872632 7976 1 0 0 10 10 0 0 0 1 0 0 211 978 695 0 0 99
is there problem on memory?
notice that po is always>0, is it normal or showing there is memory shortage? also fr is >0.
-
Looks like some paging is going on. Could you do a top and post the results. Looks like you have run out of memory. Waw, you have only 7M left. Its time to add more memory.
Sam
-
but one book i read said that,
only pi is an indicator of a memory problem.
po is normal.
based on that book, my server is ok.
who is right?
-
No. page-out shows that your system is running low in memory. When a page out occours, it goes to the disk to write the page. When the page-in occours, that means a page was requested and it was not found in memory. So page-in are common to a healthy system. Where as page outs are the indication that your system is running low in memory. To know more, do a top on your command prompt and that will tell you how much memory that you have used and how much is free, what are the process consuming how much memory, what amout of swap is used and how much is free and etc.
Sam
-
i am reading this book named "oracle high-performance tuning with staspack"
at page 95, it says
in sum, page-out operations are a normal part of virtual memory operation,but page-in(pi) operations indicate that the server have excessive RAM demands."
so who should i follow?
-
So, o.k, in simple, can you do a top and post me your top session list.
Sam
-
i am sorry, but my server don't have top installed,
can u tell me how to install it.
other machine have top though.
-
A page-in does not necessarily represent a page being recoverd from the swap area. It could be an executable code that is being paged in from a file system or could be a copy-on-write page being duplicated. Both of which are of normal occourences that do not necessarily indicate a shortage of memory. On the other hand page-out always represent data being forcibly ejected by the kernel. As I said, you can check further through $> top or vmstat -S 5.
This is the fact of the field... :) Since I haven't yet read the your book, I cannot comment on the implications.
Sam
-
in case you read that book.
please also read page 120.
at that page, the book explain why page-in indicate memory shorage, and page-out is not.
-
From the results of the vmstat, I would say your server is not even breathing hard much less working so hard it is short of RAM. Looking at the de (which will indicate short term memory deficit) and sr (which will indicate severe memory deficit if it stays high for very long).
The free memory number is misleading. In order to make the most efficient use of RAM, the system will try to continuously use up to the "lots free" setting. When it hits that limit, it will start the page daemon working to free up memory. The system will continue this activity until it hits the "min free" or "des free" parameters at which point it starts wholesale swapping of processes instead of paging. So what happens is that after a reboot you have a large number in the free column, but it slowly will approach the limit that will indicate to you what you have "lots free" set to (you can verify that free matches the system setting with your UNIX Admin -- the numbers should be very close). At that point various memory housekeeping routines attempt to keep that number stable at that point.
This particular information is specific to Sun hardware and Solaris, but most UNIX systems work in a similar method (but the parameters "min free", "lots free", and "des free" may be different). If you are on Sun and want to learn more, Adrian Cockcroft has a good book out (though it is old) called Sun Performance and Tuning, and he has columns in Sun's online magazine covering questions just like this one.
As long as scan rate stays below about 200 or 250 on 30-second intervals you should be just fine.
-
I have read that page-out will also display disk I/O that has nothing to do with paging and swapping. This becuase the kernel disk buffer is tied into the paging buffer within the kernel. That is why Sun released priority paging in 2.6 or 2.7, to increase disk I/O performance and reduce its affect on paging. In Solaris 8 you do not see disk I/O in vmstat, becuase of the new memory management. This is based on what I have been told and read.
Here is a link a friend gave me:
http://sgi-sw.rz.ruhr-uni-bochum.de/...iles/c0104.htm
[Edited by Zaggy on 07-25-2001 at 03:34 PM]
