Re: Hmmmm.

From: Mark Treacy <mark@dont-contact.us>
Date: Tue, 18 Mar 1997 18:46:47 +1100

Hi Michael,
>> Look at doing the same trick for I/O to files as per dnsserver.
>> The commercial harvest uses that with a 'filepumpd' process that
>> does the requests.
...
> #1. Something that's dead easy to do that'll give an immediate 20%
> improvement on our box is to has a seperate process doing the
> unlink()'s. i.e. spawn off
> perl -nle unlink
> and write the name of the files to the process instead of trying to do
> it in squid. Nice and easy to do.
I've had this on my todo list for a while, but have been out of round
tuits for just as long. The thing you have to watch out for is
reuse of the file after you have queued the unlink (i.e. you don't
want squid to request an unlink, reuse the file (overwrite it), and then
have the cooperating process come along and unlink it).
The simplist way to synchronise the two is to leave the file
number set in the file bitmap and defer the clear until the unlink
process acknowledges it. Whilst I don't think perl the right tool
to implement the unlink daemon, it would become
        perl -ple unlink

I think this would speed up the reload process considerably.
RebuildFromDisk would just need to temporarily set the fileno in the
filemap.

 - Mark.
Received on Mon Mar 17 1997 - 23:57:06 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:34:43 MST