Subject: | |
From: | |
Reply To: | |
Date: | Fri, 26 Mar 2004 10:18:21 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Denys writes:
< snip >
> That is what to which I took exception. I explained to Mr. Lalley that
> his understanding was faulty. XM did not scan main memory during a
> checkpoint. It only posted the pages contained in the XM log files to
> the various disk drives in the volume set being checkpointed. Now,
> whether there are one or two logfiles is totally irrelevant to that
> discussion. My point was to dispel the fantasy being perpetrated by Mr.
> Lalley and now being reinforced by Bill C.
My apologies... I was responding only to the most recent post and not to the
entire thread which I have to admit, I missed.
Yes, there are two logs per volume set, the system log and the user log, each
has two halves. One trigger for a checkpoint is when a loghalf fills. Another
trigger is when there are too many pages of XM-protected file objects frozen
in memory. There are a couple other triggers that I can't recall at the moment.
No, XM does not scan all of memory at a checkpoint. It maintains a list of
files (GUFD's) that have changed in a particular log half. A "dirty page" can
occur outside of XM, obviously, all XM cares about are the pages of objects
it has to protect.
Cheers,
Bill
===========================================
Reply to: bill . vcsy -at- comcast . net
===========================================
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
|
|