On Wed, Oct 08, 2003 at 08:08:17PM -0400, Buck wrote:

I found Hans Reiser's email address in a Google search but didn't see his site.

Hans' site is www.namesys.com Regards, -Kastus

On Thu, 2003-10-09 at 02:53, Buck wrote:

I checked out the links you sent and Reiser's website. I never made sence out of the graphs on the his site concerning the version 3 product. Version 4 looks like it must be ready to be released. It looks pretty promising. I expect I will get to taste it with whatever version follows SuSE 9. I guess for now I will stick with Reiser on the SuSE install since they at least thought enough to use it as their default.

ResierFS v4 will have many nice features like plugins. These plugins can handle things like encryption, ACL etc. I am looking forward to RFS4 for specifically this reason. I have heard they have managed to make RFS4 even faster than RFS3, so if all being well, we should be pleasantly surprised. :) Regards, -- Anders Karlsson Trudheim Technology Limited

Yes, i searched the web, My search brought about 18,100 hits most of which were links to list archives where the debate is either too technical for me to understand or its still being debated. When I received the address to his website, I checked it out. The first thing that i took notice of is that Reiser4 is not released yet. Until I today I didn't even know that there was a reiser and until after I asked the question, I didn't know there was a reiser4. Since reiser4 isn't yet released, and it is for the newer kernel than SuSE is using, that makes it irrelevant for my use. I am sorry to have cost you and the community so much of your time. I will let you go as I know you are anxious to get back to developing whatever it is that you develop. Thanks for the reply, Buck

-----Original Message----- From: rex [mailto:rex@nosyntax.net] Sent: Wednesday, October 08, 2003 8:41 PM To: suse-linux-e@suse.com Subject: Re: [SLE] Reiser vs Ext3

Buck [2003-10-08 17:10]:

...

I found Hans Reiser's email address in a Google search but didn't see his site. I emailed him directly and asked if he could give a non-technical report about what each file system advantage is and which one would he recommend.

Did you STFW? Google gives 246 hits on "reiser vs ext3" and 67 hits on a newsgroup search. A relevant message from Hans Reiser says that:

http://www.dragoninc.on.ca/mail-archives/linux-kernel/2002-10/10957.html reiser4 [2.6 kernel] is 7.6 times the write performance of ext3 for 30 copies of the Linux kernel source code using modern IDE drives and modern processors on a dual-CPU box, so I don't think any amount of improved scalability will make ext3 competitive with reiser4 for performance usages. There's a detailed paper on reiser v3 & v4 at: http://www.namesys.com/v4/fast_reiser4.html Time developers spend answering FAQs instead of coding is a loss to the community. -rex -- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Fri, 10 Oct 2003 14:53:49 -0400 Greg Engel wrote:

Here is my non-technical answer from my personal experience.

Ext3 - computer go boom, wait long time Reiser - computer go boom, comes right back up almost no waiting Why does computer go boom?

- -- Jerry Feldman Boston Linux and Unix user group http://www.blu.org PGP key id:C5061EA9 PGP Key fingerprint:053C 73EC 3AC1 5C44 3E14 9245 FB00 3ED5 C506 1EA9 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.2-rc1-SuSE (GNU/Linux) iD8DBQE/hwoj+wA+1cUGHqkRAlfgAKCHLKwXP0FLWdtgydwNoOK2JELeDQCcC4Ee Fw2oNaLG71r5tMz2ce+t+YM= =kIzd -----END PGP SIGNATURE-----

On Fri, 10 Oct 2003, Scott Jones wrote:

On Friday 10 October 2003 14:36, Jerry Feldman wrote:

...

On Fri, 10 Oct 2003 14:53:49 -0400

Greg Engel wrote:

...

Here is my non-technical answer from my personal experience.

Ext3 - computer go boom, wait long time Reiser - computer go boom, comes right back up almost no waiting

Why does computer go boom?

Power outage, bad kernel module, inquisitive small child, flying monkey invasion?

(All but one of the above have happened to me. I'll let you guess which one)

So, how did you get rid of the monkeys? -- Gerry "The lyfe so short, the craft so long to learne" Chaucer

On Saturday 11 October 2003 11:08 am, Charles Philip Chan wrote:

...

Ext3 - computer go boom, wait long time

This is not true. It comes right back up just like ReiserFS. it take less than 2 seconds to replay the journal. One thing is true though, if

Maybe the last ext I used was ext2. I can not recall when I last used it and it would have been stock SuSE.

your drive develops bad blocks, have fun recovering your data with ReiserFS!

I have never had a ReiserFS go bad, so I have never tried. Are you saying it is often not receoverable at all, or just the fsck tool used for it is difficult to use? Greg Engel

The 03.10.12 at 10:25, Greg Engel wrote:

...

your drive develops bad blocks, have fun recovering your data with ReiserFS!

I have never had a ReiserFS go bad, so I have never tried. Are you saying it is often not receoverable at all, or just the fsck tool used for it is difficult to use?

What we mean is that reiserfs does no badblock handling at all. Hard disk may have, or may develop over time, bad blocks, sectors that are not readable. Filesystems have always allowed for this, maintaining some sort of table listing those bad sectors in order not to use them. Repair tools are able to move the contents of a bad sector, if used by some file, somewhere else before marking the sector as bad. This is managed even by FAT. But reiserfs can do none of that. If you get them, you are pissed - excuse the expression. However, the current HD nowdays handles the situation in hardware. It has some sectors reserved at the factory to move there badsectors as they develop, transparently to any OS. man hdparm: -D Enable/disable the on-drive defect management fea­ ture, whereby the drive firmware tries to automati­ cally manage defective sectors by relocating them to "spare" sectors reserved by the factory for such. -- Cheers, Carlos Robinson

The 03.10.12 at 21:52, Greg Engel wrote:

So if I follow, there is a very real danger in using Reiserfs with an older drive that does not support the hardware bad sector control you are talking about.

Yes, I think so. It is not very dangerous, I think, but the danger is there. If it happens, the kernel get very slow while repeating the sector read I think ten times before giving up. If it is the root partition, then things get difficult, but it usually only means you loose one file. You need then to backup and reformat. If the sector is unrecoverable, then you need to use something different from reiserfs. Note that a bad sector, or a few, doesn't mean you have to replace the HD. Old units (10-30 Mega bytes) had a paper label listing bad sectors, hand written on the factory! -- Cheers, Carlos Robinson

The 03.10.13 at 11:06, david stevenson wrote:

As far as I know all IDE (intergrated drive electronics) and later drives do on board mapping. You need to go back to ST225 (20meg) and similar models to need user mapping.

Mapping, yes, of course. Defect mapping, I don't think so, but I could be wrong. -- Cheers, Carlos Robinson

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Monday October 13 2003 07:04, david stevenson wrote:

Have you ever seen a IDE drive shipped with a bad sector, and no they do not make them perfect, internally they all map out problems.

On Monday 13 October 2003 1:46 pm, Carlos E. R. wrote:

...

The 03.10.13 at 11:06, david stevenson wrote:

...

As far as I know all IDE (intergrated drive electronics) and later drives do on board mapping. You need to go back to ST225 (20meg) and similar models to need user mapping.

Mapping, yes, of course. Defect mapping, I don't think so, but I could be wrong.

-- Cheers, Carlos Robinson

My understanding is that the fs maps out the bad sectors and then set these not to be written to. I have been having fits of late with the reiserfs. Apprently it keeps writing to bad sectors and then causes a failure on boot. I have to rebuild the tree and then it boots. Lately this has been happening about 1 in 4 boot attempts. Of course I am investigating other reasons, since I started having these problems after updating to kde 3.1.4. But as of yet I am unable to determine exactly what is happening. Yes, the IDE drive ships with bad sectors IIRC. And the drive is mapped regardless of the drive sector status. And, as a stated before, I believe it is the fs that determines if a sectors is acceptable and then whether or not it is writable safely. If it is found bad it is marked and no data is to be written to it. Though I could be wrong in this. Does Andre Hedrik lurk on the list [:)]? I know Hans Reiser frequents SuSE bugzilla.... Hans? Cheers, Curtis. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.2-rc1-SuSE (GNU/Linux) iD8DBQE/iupxiqnGhdjCOJsRAozBAJ0TZJlPmA5eYpwM7BGkzrmzk1AwkACbB7Qa n7K3FRwrzNbKiV1Mzvf0FlQ= =CNDt -----END PGP SIGNATURE-----

Carlos E. R. wrote:

The 03.10.08 at 18:21, Buck wrote:

...

What are the advantages of using Reiser vs Ext3.

Mmmm... for example, it handles small files very easily. Or lots of files. You could have a look at the reiser web page, it explains a lot. Forgot the http...

On the other hand, I think ext3 handles badblocks better.

My personal favourite is JFS, I think it both has good performance and also good error handling. -Torvald

After taking out the http://, The computer asks me if i want to download a file. Is this url correct? Buck

-----Original Message----- From: Peter Evans [mailto:peter@despammed.com] Sent: Thursday, October 09, 2003 5:14 AM To: suse-linux-e@suse.com Subject: Re: [SLE] Reiser vs Ext3

For performance comparisons, see http://fsbench.netnation.com/ and the sparkling new Slashdot page of commentary "Linux File System Shootout".

-- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com

Thanks, I enjoyed reading that. It looks like Reiser is best for my use. Thanks again, Buck

-----Original Message----- From: Magnus Hagebris [mailto:mhagebris@ipdynamics.com] Sent: Thursday, October 09, 2003 12:23 PM To: 'Suse (E-mail)' Subject: Re: [SLE] Reiser vs Ext3

Linux Magazine has one old article talking about differences: http://www.linux-mag.com/2002-10/jfs_01.html

On Wed, 2003-10-08 at 16:23, Carlos E. R. wrote:

...

The 03.10.08 at 18:21, Buck wrote:

...

What are the advantages of using Reiser vs Ext3.

Mmmm... for example, it handles small files very easily. Or lots of files. You could have a look at the reiser web page, it explains a lot. Forgot the http...

On the other hand, I think ext3 handles badblocks better.

-- Cheers, Carlos Robinson

-- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com

On Thu, 2003-10-09 at 09:10, Kastus wrote:

On Wed, Oct 08, 2003 at 06:21:58PM -0400, Buck wrote:

...

What are the advantages of using Reiser vs Ext3.

This topic was discussed recently on suse-oracle mailing list: http://lists.suse.com/archive/suse-oracle/2003-Sep/0212.html

and according to benchmarks in http://otn.oracle.com/oramag/webcolumns/2002/techarticles/scalzo_linux02.htm... oracle DB runs more than two times faster on ext3 than on reiserfs.

This, of course, does not apply to all uses of Linux, it's just a particular case of oracle.

Reiserfs does outperform ext3 when handling small files (by a factor of) Reiserfs is geared toward regular users except if you customize (hence why ext3 will beat Reiserfs when a db is concerned is DBs use large files the opposite is true for us normal users who spend more time in the smaller file area. Reiser is a very good f.s. (but for the fs for your database yes you should consider either XFS or another fs if you want to get the last ounce of IO for those large files)

On Thursday 09 October 2003 23:34, Carlos E. R. wrote:

I understand that reiserfs is designed to be able to be... how can I explain, the database basis it self. I mean, instead of using large files and internal indexing on the database engine, move that layer to the filesystem. The reiserfs whould handle the storing and retrieving directly by records, and index it. No need to build a database on top of the filesystem and OS.

Sounds like something directly out of the IBM Mainframe world, was almost expecting to see a "Made by IBM" at the bottom. As a user, I use reiserfs. It's reliable (haven't to date encountered any problems), has working acl support, at least in SUSE. And it apparently is fast and efficient with many, small files. However, if I were on a production system. I'd put my money on the big producers, and either pick XFS or JFS, and expecting the former to be better with large databases, and the latter with distributed (medium file size) databases.

Alle 02:43, venerdì 10 ottobre 2003, Örn Hansen ha scritto:

On Thursday 09 October 2003 23:34, Carlos E. R. wrote:

...

I understand that reiserfs is designed to be able to be... how can I explain, the database basis it self. I mean, instead of using large files and internal indexing on the database engine, move that layer to the filesystem. The reiserfs whould handle the storing and retrieving directly by records, and index it. No need to build a database on top of the filesystem and OS.

Sounds like something directly out of the IBM Mainframe world, was almost expecting to see a "Made by IBM" at the bottom.

As a user, I use reiserfs. It's reliable (haven't to date encountered any problems), has working acl support, at least in SUSE. And it apparently is fast and efficient with many, small files. However, if I were on a production system. I'd put my money on the big producers, and either pick XFS or JFS, and expecting the former to be better with large databases, and the latter with distributed (medium file size) databases.

XFS would not work with gdb on suse kernels at least. (ie: gdb will return a "cannot access process memory" if you insert a breakpoint). After discovering that, all I did was replacing XFS with ReiserFS, and the latter looks a bit faster for normal use. I tested performance with bonnie and bonnie++, ReiserFS and XFS were the fastest (at the same level), then I got ext2, then ext3... and JFS was the slowest with a big gap with ext3. IIRC, bonnie stress the system with 100mbyte files. Praise

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Thursday 09 October 2003 06:34 pm, Carlos E. R. wrote:

I understand that reiserfs is designed to be able to be... how can I explain, the database basis it self. I mean, instead of using large files and internal indexing on the database engine, move that layer to the filesystem. The reiserfs whould handle the storing and retrieving directly by records, and index it. No need to build a database on top of the filesystem and OS.

I've done this before in Perl with Persistence::Object::Simple and it works well. There are a lot of neat advantages: - - Indexing is handled by the filesystem so you can perform searches with standard UNIX tools: find, locate, ls, etc - - I can modify data with a normal text editor - - No need for SQL <-> internal data structure conversion - - No need for a separate server process In Python, there's ZODB with the DirectoryStorage extension: http://dirstorage.sourceforge.net/ - -- James Oakley Engineering - SolutionInc Ltd. joakley@solutioninc.com http://www.solutioninc.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.2-rc1-SuSE (GNU/Linux) iD8DBQE/hsE5+FOexA3koIgRArDHAKCrssVJSzrZRiYqhz//760lt2wooQCePgRW V1E7KTjFZCGwlnzAZWNCfxo= =raM6 -----END PGP SIGNATURE-----

On Thursday 09 October 2003 02:10, Kastus wrote:

On Wed, Oct 08, 2003 at 06:21:58PM -0400, Buck wrote:

...

What are the advantages of using Reiser vs Ext3.

This topic was discussed recently on suse-oracle mailing list: http://lists.suse.com/archive/suse-oracle/2003-Sep/0212.html

and according to benchmarks in http://otn.oracle.com/oramag/webcolumns/2002/techarticles/scalzo_linux02.ht ml oracle DB runs more than two times faster on ext3 than on reiserfs.

Oracle, is not made of many small files ... for such use, something Like XFS, or JFS would be appropriate. Ext3 is a journal add-on, to Ext2 and backwards compatible as far as I can understand. It's based on the legacy unix file system, which is really out of date.

On Thursday 09 October 2003 23:00, Gerhard den Hollander wrote:

It's been around longer then Reiserfs, and has been in much more use then reiser, which menas that any bugs have a higher likely hood of being already found (and fixed) on ext3 then on reiserfs.

But Reiser has been around long enough for horror stories to surface if they were going to. Its largely been trouble free. SuSE has pretty well standardized on it, and you just don't see many complaints. When is long enough? -- _____________________________________ John Andersen

Graham Smith wrote:

Introduction to Linux filesystems and files

http://newsforge.com/newsforge/03/10/07/196222.shtml?tid=23 Very intereseting with some correction and thoughts about a journaling FS..

As I have heard the MS NTFS is a rather poor effort at a journaling file system and that mounting a NTFS is read only for security reasons. I have heard that if it is set to write you can write to the NTFS FS BUT! failure is likely. This is a issue with the journaling NTFS losing it mind trying to figure out what happened when booted up. With Linux Reiserfs and ext3 I have never seen this same kind of problem. This article implies that reiserfs could have a similar problem. Have I just missed it? -- 73 de Donn Washburn __ " http://www.hal-pc.org/~n5xwb " Ham Callsign N5XWB / / __ __ __ __ __ __ __ 307 Savoy St. / /__ / / / \/ / / /_/ / \ \/ / Sugar Land, TX 77478 /_____/ /_/ /_/\__/ /_____/ /_/\_\ LL# 1.281.242.3256 a MSDOS Virus "Free Zone" OS Email: n5xwb@hal-pc.org Info: http://www.knoppix.net

Just what the doctor ordered! Thank you, Graham! :) Buck

-----Original Message----- From: Graham Smith [mailto:gqs@iinet.net.au] Sent: Friday, October 10, 2003 11:03 PM To: suse-linux-e@suse.com Subject: Re: [SLE] Reiser vs Ext3

Introduction to Linux filesystems and files

http://newsforge.com/newsforge/03/10/07/196222.shtml?tid=23 -- Regards,

Graham Smith ---------------------------------------------------------

-- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com

On Fri, 10 Oct 2003 20:59:22 -0700 Art Fore wrote:

I find Reiser far superior to Ext3, especially with power outages which are not too infrequent in my area. With Ext3, it would take 10 or 20 minutes, with Reiser, you hardly notice it.

Something is wrong here, it should only take ext3 less than 2 seconds to replay the journal, So, either your drive is not has really ext3 or it is still being mounted as ext2. Charles -- "I once witnessed a long-winded, month-long flamewar over the use of mice vs. trackballs...It was very silly." (By Matt Welsh)

On Saturday 11 October 2003 03:51 am, John Andersen wrote:

But Reiser has been around long enough for horror stories to surface if they were going to. Its largely been trouble free. SuSE has pretty well standardized on it, and you just don't see many complaints.

I still find it impossible to maintain a stable nfs setup when the exported filesystem is reiser - no matter how much people say the problems are solved. Dylan -- Sweet moderation Heart of this nation Desert us not We are between the wars - Billy Bragg

On Saturday 11 October 2003 18:05 pm, Buck wrote:

What file system do you find fits that purpose best?

Well, until I hear convincing evidence otherwise, I'm sticking to ext3. With reiser as an exported fs, after a while (usually a week or so) the clients start getting progressively slower on nfs mounts, and file operations (including directory listings) until timeouts start kicking in and operations fail. At that point, I start getting RPC: permission denied etc, ant have to manually clear out the nfs files in /var to allow a day of faultless function. Also, autofs simply refuses to play ball. With ext3 (and ext2) I have had no issues at all.

Is Reiser good for you on the partitions that aren't shared?

I don't know - for simplicity (and to remove the need for initrd modules) I use ext3 throughout. I don't see the point of mixing filesystems as I have no pressing need for the specialities of any of the others. Dylan -- Sweet moderation Heart of this nation Desert us not We are between the wars - Billy Bragg

On Saturday 11 October 2003 18:53 pm, js wrote:

This is wierd. All of my nfs mounts are reiser. I have never had any of the problems you mention. And like one of the earlier posts, I've noticed it takes much longer to play logs on ext3 which is the sole reason I stopped using it months ago. YMMV.

Indeed. For me, ext3 log replay takes seconds - but then, the periodic full checks at reboot can take ages Dylan -- Sweet moderation Heart of this nation Desert us not We are between the wars - Billy Bragg

On Sat, 11 Oct 2003 19:05:04 +0100 Dylan wrote:

Indeed. For me, ext3 log replay takes seconds - but then, the periodic full checks at reboot can take ages

The checks are a good thing- a journaling FS does not really guarantee all data is secure (just meta-data). If you don't like it you can either increase the "maximum mount count" or disable it totally with tune2fs. Charles -- "If you want to travel around the world and be invited to speak at a lot of different places, just write a Unix operating system." (By Linus Torvalds)