Re: dirpref gives massive performance boost
On 2001-Oct-27 21:48:38 -0500, Glenn Johnson [EMAIL PROTECTED] wrote: I figured out what I did wrong. In the instructions above it says to edit 'etc/fstab'; I edited '/etc/fstab' instead. A very easy mistake to make. It can get very confusing when you have multiple, almost-identical partitions mounted. I'm glad that you sorted it out Also, the line in the instructions that says '0,ad(0,b)/boot/loader' should be '0:ad(0,b)/boot/loader' That's what I get for writing from memory. Luckily boot0 reports the syntax so hopefully this didn't cause you too much confusion. Peter To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
Tim Bunce [EMAIL PROTECTED] wrote: Can the diprefs code have a useful effect on an individual subtree of a file system if just that tree was deleted and recreated? Well, it depends on how much free space there is on the filesystem, and how fragmented it is. If the filesystem is 90% used, the dirprefs code doesn't have much room to use disk blocks for new directories in an efficient way. But it's probably better than nothing. It's very difficult to say in advance, so I'd suggest you just try it. Regards Oliver -- Oliver Fromme, secnetix GmbH Co KG, Oettingenstr. 2, 80538 München Any opinions expressed in this message may be personal to the author and may not necessarily reflect the opinions of secnetix in any way. All that we see or seem is just a dream within a dream (E. A. Poe) To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
Robert Gray [EMAIL PROTECTED] wrote: A related question that I was wondering, Does my /kernel have the new dirpref code? If you've build kernel and userland from the same date (which is always recommended), then it is sufficient to look at the newfs(8) manpage. If it has the -g and -h options, then you've got dirprefs in the kernel. They were introduced at the same time as the kernel dirprefs code. Regards Oliver -- Oliver Fromme, secnetix GmbH Co KG, Oettingenstr. 2, 80538 München Any opinions expressed in this message may be personal to the author and may not necessarily reflect the opinions of secnetix in any way. All that we see or seem is just a dream within a dream (E. A. Poe) To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
On Thu, Oct 18, 2001 at 12:32:22PM -0500, David W. Chapman Jr. wrote: If it's a UFS, and you're running a new enough kernel, it's getting done. Of course, it can only be smart about stuff written to the disk after you built the new kernel. If your disk is mostly empty, you're cool. If it's pretty full you should back it up, delete everything (newfs is the fastest way) and then restore all your data. Well yes, but I would like to avoid all of this if it is already using dirpref. I currently don't have enough drive space to fully backup my 24gig stripe set. Any directories created after you installed a kernel with the new filesystem layout code will have the more optimized layout. Directories created earlier will have the layout they've always had. Since there's not a switch to turn on/off or test, there's no good way to tell how close to optimal your current disk is. You can force it to be nice by dumping and restoring it. Otherwise, I'd say the create date/time stamp compared against a calendar would give you your best guess. -crl -- Chad R. Larson (CRL15) 602-953-1392 Brother, can you paradigm? [EMAIL PROTECTED] [EMAIL PROTECTED] [EMAIL PROTECTED] DCF, Inc. - 14623 North 49th Place, Scottsdale, Arizona 85254-2207 To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
On Thu, Oct 18, 2001 at 08:26:00PM +0100, Tim Bunce wrote: Can the diprefs code have a useful effect on an individual subtree of a file system if just that tree was deleted and recreated? Sure. Do a rm -rf /usr/ports/* and then re-sup the damn thing. You'll be impressed. -crl -- Chad R. Larson (CRL15) 602-953-1392 Brother, can you paradigm? [EMAIL PROTECTED] [EMAIL PROTECTED] [EMAIL PROTECTED] DCF, Inc. - 14623 North 49th Place, Scottsdale, Arizona 85254-2207 To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
At 10:57 AM -0500 10/18/01, David W. Chapman Jr. wrote: Must one supply any other arguments to newfs in order to enable dirpref? A quick look at man newfs didn't make any mention of dirpref. No, it's on by default in kernels that include the new code. Is there a way to check to see if a slice has difpref enabled? Dirpref is not something which is enabled or disabled, not in the same sense as softupdates is enabled. Dirpref is a smarter layout of information in a partition. You need a version of the system which knows HOW to do that smarter layout, and then you just rebuild the partition. There is no switch to turn on and off. -- Garance Alistair Drosehn= [EMAIL PROTECTED] Senior Systems Programmer or [EMAIL PROTECTED] Rensselaer Polytechnic Instituteor [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
On Thu, Oct 18, 2001 at 02:48:50PM +0100, Rasputin wrote: * Doug Poland [EMAIL PROTECTED] [011018 14:40]: On Tue, Oct 02, 2001 at 09:43:53PM +0200, Erik Trulsson wrote: On Tue, Oct 02, 2001 at 09:20:02PM +0200, Steve O'Hara-Smith wrote: On Fri, 28 Sep 2001 14:12:46 -0700 Kris Kennaway [EMAIL PROTECTED] wrote: KK Hi all, KK KK Just a note to those who have updated to 4.4-STABLE that it's well KK worth doing a backup + newfs + restore on all your UFS volumes. The CAVEAT: insert a tunefs -n enable between newfs and restore, without it the performance improvement is rather less dramatic :) You can happily skip the tunefs step if you just remember to supply the '-U' flag to newfs. Must one supply any other arguments to newfs in order to enable dirpref? A quick look at man newfs didn't make any mention of dirpref. No, it's on by default in kernels that include the new code. Is there a way to check to see if a slice has difpref enabled? -- David W. Chapman Jr. [EMAIL PROTECTED] Raintree Network Services, Inc. www.inethouston.net [EMAIL PROTECTED] FreeBSD Committer www.FreeBSD.org To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
Dirpref is not something which is enabled or disabled, not in the same sense as softupdates is enabled. Dirpref is a smarter layout of information in a partition. You need a version of the system which knows HOW to do that smarter layout, and then you just rebuild the partition. There is no switch to turn on and off. I'm not looking to turn if off or on, just to see whether a file system of mine has that capability in it or if I need to newfs it. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
In message [EMAIL PROTECTED], Mike Harding writes: Um - what about my question? Is the newfs necessary? Repartitioning, backup, and restore also require a backup medium, etc. The more of your filesystem that has had its files allocated using the new dirpref, the greater the benefit. There was a comment made earlier in another thread on another list (see archives) that the new dirpref has the risk of greater fragmentation. There were no responses to the comment. I'm not sure whether this is just a concern someone had or whether the risk is real. The fact that there were no replies to that comment seems to indicate we just don't know yet. In some OpenBSD mailing list archives (search Google) there were comments about dirpref + softupdates being 60x faster than UFS without the two features. Regards, Phone: (250)387-8437 Cy SchubertFax: (250)387-5766 Team Leader, Sun/Alpha Team Internet: [EMAIL PROTECTED] Open Systems Group, ITSD Ministry of Management Services Province of BC To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: dirpref gives massive performance boost
So it sounds like there would be some benefit in tar'ing and untarring /usr/local, /usr/ports, /usr/src, etc. which will be less disruptive... Thanks! - Mike H. X-Authentication-Warning: cwsys.cwsent.com: smtpd set sender to cy@cwsys using -f Reply-To: Cy Schubert - ITSD Open Systems Group [EMAIL PROTECTED] From: Cy Schubert - ITSD Open Systems Group [EMAIL PROTECTED] X-Sender: schubert Cc: [EMAIL PROTECTED], [EMAIL PROTECTED] Content-Type: text/plain; charset=us-ascii Date: Sun, 30 Sep 2001 08:49:47 -0700 Sender: [EMAIL PROTECTED] X-SpamBouncer: 1.4 (8/24/01) X-SBClass: OK In message [EMAIL PROTECTED], Mike Harding writes: Um - what about my question? Is the newfs necessary? Repartitioning, backup, and restore also require a backup medium, etc. The more of your filesystem that has had its files allocated using the new dirpref, the greater the benefit. There was a comment made earlier in another thread on another list (see archives) that the new dirpref has the risk of greater fragmentation. There were no responses to the comment. I'm not sure whether this is just a concern someone had or whether the risk is real. The fact that there were no replies to that comment seems to indicate we just don't know yet. In some OpenBSD mailing list archives (search Google) there were comments about dirpref + softupdates being 60x faster than UFS without the two features. Regards, Phone: (250)387-8437 Cy SchubertFax: (250)387-5766 Team Leader, Sun/Alpha Team Internet: [EMAIL PROTECTED] Open Systems Group, ITSD Ministry of Management Services Province of BC To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message