In the days of old when swap size was swap size and where everything in
memory had a copy in the swap space, that space actually set the limit for
"real space". In effect, virtual space was exactly the size of the swap
space, so you'd set that size to some increment over the real memory size.
Some folks like a 2:1 overcommitment ratio where the swap space is twice
the size of real memory (let's leave out the size of the OS and buffer
cache, OK?)
With demand paging and block loading, the old rules of thumb for sizing
swap space need to be re-thought. In effect, for the executable ("pure"
code, in the "code segment") code, the actual file being executed is part
of the paging space.
First off, your workload will have a point where it will thrash at a
specific overcommitment ratio. This crossover point is specific to a given
workload.
Total virtual size in a paging system is the real memory + paging space
(and, for linux: + size of code segments of all executing programs).
While you are likely under some pressure to shring your Linux instance's
memory footprint, you need to take a look at how much memory is actually in
use and for what-- i.e. real data (what I recall AIX referring to as
"computational space") and I/O buffers.
Additionally-- and here's one of the debatable points-- making your
instance bigger and doing away with paging space entirely may be very
tempting, allowing z/VM to manage actual paging, except, of course, that
you then have to tune Linux's buffer allocation sizing to minimize the
space allocated (as it currently stands, Linux will, if the workload is
"right", often try to use all of the memory not busy carrying a program's
code or data segments as I/O buffer space).
Otherwise, it strikes me that this is one of those subjects that can drive
yet another religious war...
--------------------
John R. Campbell, Speaker to Machines (GNUrd) (813) 356-5322 (t/l 697)
Adsumo ergo raptus sum
MacOS X: Because making Unix user-friendly was easier than debugging
Windows.
Red Hat Certified Engineer (#803004680310286)
IBM Certified: IBM AIX 4.3 System Administration, System Support
----- Forwarded by John Campbell/Tampa/IBM on 07/26/06 12:48 PM -----
Brian France
<[EMAIL PROTECTED]>
Sent by: Linux on To
390 Port [email protected]
<[EMAIL PROTECTED] cc
IST.EDU>
Subject
[LINUX-390] Swap size to Memory
07/26/06 11:39 AM size...
Please respond to
Linux on 390 Port
<[EMAIL PROTECTED]
IST.EDU>
Folks,
I have an image that is 1280m. It's swap space is 464m. Is that a
good ratio? The image has chewed up according to Perftoolkit 75% of
it. I was thinking a bigger swap space as opposed to more memory
which is what I being pressured to do. Response times seems great.
This image is running DB2 and Tamino (SAG data base) and some other
things. I just was sure if there was a good rule of thumb to follow
regarding image size and swap space. THANX!!!
Brian W. France
Systems Administrator (Mainframe)
Pennsylvania State University
Administrative Information Services - Infrastructure/Sysarc
Rm 25 Shields Bldg., University Park, Pa. 16802
814-863-4739
[EMAIL PROTECTED]
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
