Re: Management of Samba (again)

David Collier-Brown (davecb@Canada.Sun.COM)
Mon, 30 Mar 1998 08:19:14 -0500

Date:	Mon, 30 Mar 1998 08:19:14 -0500
From:	David Collier-Brown <davecb@Canada.Sun.COM>
To:	lkcl@switchboard.net
Subject: Re: Management of Samba (again)

Luke Kenneth Casson Leighton wrote:
> ok, let me say that again. you got my support for "modularisation" if it
> also involves cacheing of the parameter configuration, in each of the smbd
> daemons.

This was done successfully in Rayan Zachariasson's zmailer:
what was notable there was the clean characterization of the
thing being cached.

Unsorted flat files had a medium-sized cache, and an
inexpensive test for staleness (stat ::= read an inode)
In Samba's case, the cache subsumes the whole file (:-))

Sorted flat files and dbx had such a low cost to search
(one to two seek/read pairs) that the cache in the program
could be very small (just the last thing read, if my
leaky memory suffices), and

Daemons with long times to check or lookup had relatively
large caches, with explicit ``times to live'' (eg, DNS,
any browsing scheme).

This gave very good performance in the ``usual undesirable case'',
that of looking up the same thing repeatedly in umpty-dozen
servers, and good performance in the general case.

--dave

-- 
David Collier-Brown,  | Always do right. This will gratify some people
185 Ellerslie Ave.,   | and astonish the rest.        -- Mark Twain
Willowdale, Ontario   | davecb@hobbes.ss.org, canada.sun.com
M2N 1Y3. 416-223-8968 | http://java.science.yorku.ca/~davecb