Date: Mon, 25 Jan 1999 20:18:12 -0500 (EST) From: Brian Feldman <green@unixhelp.org> To: Matthew Dillon <dillon@apollo.backplane.com> Cc: Sheldon Hearn <axl@iafrica.com>, Lee Cremeans <lee@st-lcremean.tidalwave.net>, current@FreeBSD.ORG Subject: Re: IDE DMA works, I'll be a... Message-ID: <Pine.BSF.4.05.9901252016180.7918-100000@janus.syracuse.net> In-Reply-To: <199901260059.QAA08899@apollo.backplane.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 25 Jan 1999, Matthew Dillon wrote:
> :On Mon, 25 Jan 1999 19:40:38 EST, Brian Feldman wrote:
> :
> :> I say this because I am down to ~2.5MB/s on each hard drive, a MB or
> :> so decrease; I also seem to be reading (bs=512k) from the CD uncooked
> :...
> :
> :Since I rebuilt world shortly after Matt's VM surgery started, I've also
> :noticed this -- lower apparent transfer rates from /dev/zero to disk,
> :higher CPU usage during such transfers.
> :
> :I'm _not_ saying it's Matt's stuff that's done this, it's just that that
> :was the stuff I _noticed_ happening in commit mail. I think I saw
> :something happening to wd.c recently as well, so I _definitely_ am not
> :going on record as laying blame.
> :
> :Just some confirmation. :-)
> :
> :Ciao,
> :Sheldon.
>
> Hmm, interesting. A dd copy doesn't touch on the main of what I did,
> but I do have a (bad) hack in there to handle a case that John brought
> up in regards to the inactive page queue getting overloaded with
> clean meta-data. It might be interesting to turn that off and see if
> your transfer rates improve.
>
> You can turn off the hack by commenting out the case that activates
> it: Add the #if/#endif and split the open brace out from the else,
> as shown.
>
> I would like to know if your transfer rate improves or not, and by
> how much.
>
>
> /*
> * Figure out what to do with dirty pages when they are encountered.
> * Assume that 1/3 of the pages on the inactive list are clean. If
> * we think we can reach our target, disable laundering (do not
> * clean any dirty pages). If we miss the target we will loop back
> * up and do a laundering run.
> */
>
> #if 0
> if (cnt.v_inactive_count / 3 > page_shortage) {
> maxlaunder = 0;
> launder_loop = 0;
> } else
> #endif
> {
> maxlaunder =
> (cnt.v_inactive_target > max_page_launder) ?
> max_page_launder : cnt.v_inactive_target;
> launder_loop = 1;
> }
>
>
> -Matt
> Matthew Dillon
> <dillon@backplane.com>
>
I'd rather use if (0 && cnt.v_inactive_count / 3 > page_shortage) {. I'm
trying it out now, here's the "before":
IOZONE performance measurements:
2972048 bytes/second for writing the file
2962863 bytes/second for reading the file
>
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-current" in the body of the message
>
Brian Feldman _ __ ___ ___ ___
green@unixhelp.org _ __ ___ | _ ) __| \
http://www.freebsd.org/ _ __ ___ ____ | _ \__ \ |) |
FreeBSD: The Power to Serve! _ __ ___ ____ _____ |___/___/___/
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.05.9901252016180.7918-100000>