swap log

From: tk <tk@dont-contact.us>
Date: Wed, 28 Nov 2001 23:47:14 +0100

Hello all,

I was discussing the Stefan R van den Berg patches with stefan and
directed my project toward a patch for the coss file system. I need some
help, I'm having difficulties understanding some things. Here is the
idee of the patches:

One part of the patch is the problem of sequentialy store the swap log
file. The idea was to reduce (in some cases eleminate) the growth of
this file.
- Instead of add a delete record to the swaplogfile, put a zero (empty
marker) on the first bit of the record, and fill the next 4 bytes with a
pointer to the next empty-list. If a record is deleted the last
empty-list is filled with a pointer pointing to this deleted record and
the last pointer in the chain is NULL.All record containing data get a 1
on the first bit.
If an empty-list pointer points to a NULL means that the file is empty
on deleted records, so it had to create a new record in the tail of the
swaplogfile.This will reduce the start up time of squid in case of failure.
My question is: Is memory mapping this swap log file a speed
improvement? Will this modyfing the swap log file really be a speed
improvement on startup ?

Thank you for your time,
Tom Otermans, Student Technical Computer Sience
Received on Wed Nov 28 2001 - 15:55:39 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:14:39 MST