If you are talking about what is actually residing on the swap device, then
I would agree with you.  And from the perspective of z/VM and VDISK, you
would also be right.  But the question was asking about what *Linux* thought
was out there.  The idea was that Linux thought that 100 paging slots were
*in use* and could not have other pages written in to them.  I still say
that would not be the case.  If you could examine the control blocks, Linux
would think those 99 other slots were available for use.


Mark Post

-----Original Message-----
From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Barton
Robinson
Sent: Monday, February 28, 2005 4:55 PM
To: [email protected]
Subject: Re: Anyone running WebSphere under Linux on z/Series AND using ws
admin


 scripts for application deployment?

Sorry Mark, I would have to say YES, there are 99 stale copies.... I assume
these are virtual disks? And you measure from z/VM, and not Linux?  If
measuring from Linux and the swap full goes up, then you have a "swap
cancer" and that is bad.

Linux uses a moving cursor algorithm for swap space allocation. You have
done the right thing by having 3 swap disks, prioritized. I think it was Rob
van der Heij that came up with this technique. BUT, they may be - or likely
are too large. The moving cursor algorithm will allocate all of the first
disk, and then go back and re-use it until it is full, then ONLY when the
first disk is full, the second disk will be utilized. Thus the optimum
configuration is for the first swap disk to be the correct size at peak
period. The second disk is just for occassional overflow.  If the first disk
is too large, then you WILL have 99 stale copies.  If sized correctly, then
linux will write over the old ones.

Look at the Linux data to understand the maximum that Linux thinks it has
allocated as swap. That is the size of the first virtual disk....

>I would have to say no, you don't have 99 stale copies taking up paging
>slots.
>
>
>Mark Post
>
>-----Original Message-----
>From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of
>James Melin
>Sent: Friday, February 25, 2005 11:45 AM
>To: [email protected]
>Subject: Re: Anyone running WebSphere under Linux on z/Series AND using
>wsadmin scripts for application deployment?
>
>
>From what I've read this morning, I think I am being victimized by the
>2.4 kernel design, as far as swapping is concerned.
>
>If I have 3 swap volumes:
>
>pequot:~ # swapon -s
>Filename                        Type            Size    Used    Priority
>/dev/dasde1                     partition       1201576 0       -1
>/dev/dasde2                     partition       1202392 0       -2
>/dev/dasda1                     partition       801256  0       -3
>
>When will the operating system attempt to clean up swap? When all three
>swap spaces are filled? If I swap out a page, swap it in again, swap it
>out again, and in again... lets say I do this 100 times.... do I have
>99 'stale' copies of the page on the swap volume?







"If you can't measure it, I'm Just NOT interested!"(tm)

/************************************************************/
Barton Robinson - CBW     Internet: [EMAIL PROTECTED]
Velocity Software, Inc    Mailing Address:
 196-D Castro Street       P.O. Box 390640
 Mountain View, CA 94041   Mountain View, CA 94039-0640

VM Performance Hotline:   650-964-8867
Fax: 650-964-9012         Web Page:  WWW.VELOCITY-SOFTWARE.COM
/************************************************************/

----------------------------------------------------------------------
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

Reply via email to