Date: Thu, 13 Mar 1997 10:59:12 +0900 (JST) From: Michael Hancock <michaelh@cet.co.jp> To: Joe Greco <jgreco@solaria.sol.net> Cc: Tom Samplonius <tom@sdf.com>, taob@risc.org, mmead@goof.com, isp@freebsd.org, hackers@freebsd.org Subject: block and frag size for news (was Re: freebsd as a news server?) Message-ID: <Pine.SV4.3.95.970313104650.24048C-100000@parkplace.cet.co.jp> In-Reply-To: <199703111608.KAA28112@solaria.sol.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 11 Mar 1997, Joe Greco wrote: > Yeah, well, I've traditionally used 4096/512 but in the last year or so > it's seemed to me that a few machines that I have with 8192/1024 are > "faster" (based mostly on feel). I had done some experimentation that > seemed to support that. I'm using 4096/1024. I think the overhead of managing 512K frags is not worth the efficient use of space and this is probably the performance problem. I would think that having a block size at 8192 would cause too many frag to block promotions on a news spool. Again I've been using. newfs -i 3072 -b 4096 -f 1024 -a 8 I used -a 8 to leave it at the default and as Bruce points out since the rotational delay now defaults to 0 the maxcontig setting probably doesn't matter. It might matter I suppose if it actually goes thru the motion of triggering a rotational delay, even if it's 0, after every 8 blocks. Using -i 3072 makes my df and df -i %Free columns very close for most spools. Regards, Mike Hancock
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.SV4.3.95.970313104650.24048C-100000>