* What Filesystem?
@ 2003-01-29 20:20 James Thompson
2003-01-29 15:28 ` Ross Vandegrift
` (2 more replies)
0 siblings, 3 replies; 10+ messages in thread
From: James Thompson @ 2003-01-29 20:20 UTC (permalink / raw)
To: reiserfs-list
Dear Sirs,
I am a visual artist and musician. As well as a range of traditional
media I use Linux Mandrake 8.0 on an old Athlon-based PC. I wrote to
Mr. Hans Reiser, after following a mailto link on the Namesys website,
with a slightly lengthier version of the question written below, and was
advised that "This is really a better question for our mailing list
(reiserfs-list@namesys.com) than for me, as I am obviously biased." I
am therefore sending this email to you in the hope that somebody may be
able to solve my puzzle - or at least direct me towards another
knowledgeable source of information! Thank you.
>> The question:
... I have done much research in the field of computer /hardware/
suitable for commercial Digital Content Creation (P4 Xeon; Wildcat
graphics; ART's "PURE" raytracing PCI render board, etc. ...) and now
look to a better understanding of the choices I might make for a
(presumably) Linux-based OS running on (presumably) Intel 32-bit
hardware SUCH AS these two elements:-
1. Kernel Patches - pre-empt and low latency;
2. File System type - EXT3, ReiserFS, SGi's XFS, HFS, JFS....;
Would you be able to advsie me on any issues I might need to be aware of
and perhaps any firm decisions you think would be good for me to make,
regarding OS choice and configuration? I realise that all pre-built
workstations supplied by Dell, IBM and HP-Compaq come with RedHat, but,
just for the record, I like Mandrake and I am getting into the
WindowMaker desktop :).
I also develop my website locally using Apache (but hosted on Freeserve
:( ), and I understand pre-empt &/or low-latency patches are
counter-effective for servers. However, as long as no real harm can be
expected, my priority is for graphics and sound creation, editing,
compositing, publishing, etc., while fast response for just developing
html pages isn't.
I would like to believe that, somewhere, I can get a 'standard' Linux
patched for optimal DCC, with suitable FS type available to choose from
during installation. If not, I will have to find and apply patches, but
I know that with FS type any change demands a reformat, which can really
disrupt a working week!
>> [End question]
Yours faithfully,
James Thompson,
Visual Artist & Musician
H.E. Student of Fine Arts
^ permalink raw reply [flat|nested] 10+ messages in thread* Re: What Filesystem? 2003-01-29 20:20 What Filesystem? James Thompson @ 2003-01-29 15:28 ` Ross Vandegrift 2003-01-29 15:32 ` Anders Widman 2003-01-29 19:02 ` Dieter Nützel 2003-01-29 15:58 ` Sam Vilain 2003-01-29 18:16 ` Valdis.Kletnieks 2 siblings, 2 replies; 10+ messages in thread From: Ross Vandegrift @ 2003-01-29 15:28 UTC (permalink / raw) To: James Thompson; +Cc: reiserfs-list On Wed, Jan 29, 2003 at 03:20:26PM -0500, James Thompson wrote: > I am a visual artist and musician. Check out the document at http://myweb.cableone.net/eviltwin69/ALSA_JACK_ARDOUR.html. There a section that benchmarks various filesystems for their latency. The short story is that Reiserfs wins handily over Ext2, Ext3, and FAT32 (duh! why would someone test that...). Unfortunately, there's no comparison between Reiser/JFS/XFS. I think that would be more of a fair match. Anyhow though, for general low-latency multimedia work, ReiserFS looks like it's a good choice. -- Ross Vandegrift ross@willow.seitz.com A Pope has a Water Cannon. It is a Water Cannon. He fires Holy-Water from it. It is a Holy-Water Cannon. He Blesses it. It is a Holy Holy-Water Cannon. He Blesses the Hell out of it. It is a Wholly Holy Holy-Water Cannon. He has it pierced. It is a Holey Wholly Holy Holy-Water Cannon. He makes it official. It is a Canon Holey Wholly Holy Holy-Water Cannon. Batman and Robin arrive. He shoots them. ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 15:28 ` Ross Vandegrift @ 2003-01-29 15:32 ` Anders Widman 2003-01-29 19:02 ` Dieter Nützel 1 sibling, 0 replies; 10+ messages in thread From: Anders Widman @ 2003-01-29 15:32 UTC (permalink / raw) To: reiserfs-list > On Wed, Jan 29, 2003 at 03:20:26PM -0500, James Thompson wrote: >> I am a visual artist and musician. > Check out the document at > http://myweb.cableone.net/eviltwin69/ALSA_JACK_ARDOUR.html. There a > section that benchmarks various filesystems for their latency. The > short story is that Reiserfs wins handily over Ext2, Ext3, and FAT32 > (duh! why would someone test that...). Simply because FAT filesystem is very simple and has little overhead. If dealing with little amount of files and folders then it can be rather quick, especially if you need CPU resources to other things. > Unfortunately, there's no > comparison between Reiser/JFS/XFS. I think that would be more of a fair > match. > Anyhow though, for general low-latency multimedia work, ReiserFS looks > like it's a good choice. ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 15:28 ` Ross Vandegrift 2003-01-29 15:32 ` Anders Widman @ 2003-01-29 19:02 ` Dieter Nützel 1 sibling, 0 replies; 10+ messages in thread From: Dieter Nützel @ 2003-01-29 19:02 UTC (permalink / raw) To: Ross Vandegrift, James Thompson; +Cc: reiserfs-list Am Mittwoch, 29. Januar 2003 16:28 schrieb Ross Vandegrift: > On Wed, Jan 29, 2003 at 03:20:26PM -0500, James Thompson wrote: > > I am a visual artist and musician. > > Check out the document at > http://myweb.cableone.net/eviltwin69/ALSA_JACK_ARDOUR.html. There a > section that benchmarks various filesystems for their latency. The > short story is that Reiserfs wins handily over Ext2, Ext3, and FAT32 > (duh! why would someone test that...). Unfortunately, there's no > comparison between Reiser/JFS/XFS. I think that would be more of a fair > match. XFS could win _today_ for "real time" (granted video bandwidth, for which it was designed in the first place), but this could change (compensate) with the latest ReiserFS 3.x patches (data logging) and finally Reiser4 (see the Reiser Homepage). > Anyhow though, for general low-latency multimedia work, ReiserFS looks > like it's a good choice. Yes. Go with low-latency _and_ preemption patches (try Gentoo, it has it all). I did beta testing for Robert Love (MontaVista) for ages and it is _the way to go_ for multi media machines. Greetings, Dieter -- Dieter Nützel Graduate Student, Computer Science University of Hamburg Department of Computer Science @home: Dieter.Nuetzel at hamburg.de (replace at with @) ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 20:20 What Filesystem? James Thompson 2003-01-29 15:28 ` Ross Vandegrift @ 2003-01-29 15:58 ` Sam Vilain 2003-01-29 19:31 ` Dieter Nützel 2003-01-29 18:16 ` Valdis.Kletnieks 2 siblings, 1 reply; 10+ messages in thread From: Sam Vilain @ 2003-01-29 15:58 UTC (permalink / raw) To: james, reiserfs-list Reiserfs is probably not what you want for doing lots of high volume data. Reiserfs is good with many small files and general purpose use. It is actually the slowest filesystem for bulk data. I'm sure some smartass will probably post some benchmark to prove me wrong, but SGI have a long heritage of making filesystems for exactly what you want and so if XFS works I'd say use that. It's pumpin'. Of course more quantitative comparisons can easily be found on the linux-kernel mailing list. However, the practical difference between the high performance filesystems you mentioned, or the difference between running pre-empt or not should be considered marginal compared to other factors - such as, the load that your hard disk controller places on the system and the number of physical disks you have (and how many RPM they run at). Keep in mind that you can often nearly double the data throughput of a system by doubling the number of physical disks in it, and using RAID. Even with the latest UDMA-133, I haven't seen any IDE system actually perform without bothering the CPU non-trivially (of course YMMV). Using SCSI disks and controllers will give you a smoother system ride; which is why 95% of high-end workstations come equipped with SCSI. You can get 15,000 RPM U320 SCSI disks, which are f*** fast (though loud). This is largely because the SCSI protocol was designed properly, IDE/ATA is a hack. Serial-ATA promises to offer SCSI like host efficiency, but I'll only believe it when I see it. And at the moment the costs are as bad as SCSI anyway. Of course, being able to move around large chunks of data quickly extends to other parts of the system, too. The bigger and faster the system BUS, the better. Having `researched hardware platforms' you should know this, of course. As far as Athlon Motherboards go, Tyan are a reputable vendor who used to produce Sun clone motherboards - they have a really nice dual capable system with dual U160 SCSI controllers and 66MHz PCI slots - which must mean that the SCSI controllers run at that speed. It's definitely worth the ~$500 price tag. They all pale in comparison to R10000+ based SGI or Sparc 9+ platforms of course. From a graphic artist's perspective, you're probably better off buying a new G4 based system and running MacOS X, you know :-). Linux isn't exactly `The Platform' for digital content creation. The Mac's CPU and hardware platform are a lot better at moving data around, and if you've got a Mac then you can run Mac OS X, or Linux + Mac OS inside an emulator (which runs FAST!). As a final note, keep in mind that a filesystem reformat does not mean a re-install; keep your partitions as small as practical and you can change them over individually. Sam. On Thu, 30 Jan 2003 09:20, James Thompson wrote: > Dear Sirs, > > I am a visual artist and musician. As well as a range of traditional > media I use Linux Mandrake 8.0 on an old Athlon-based PC. I wrote to > Mr. Hans Reiser, after following a mailto link on the Namesys website, > with a slightly lengthier version of the question written below, and was > advised that "This is really a better question for our mailing list > (reiserfs-list@namesys.com) than for me, as I am obviously biased." I > am therefore sending this email to you in the hope that somebody may be > able to solve my puzzle - or at least direct me towards another > knowledgeable source of information! Thank you. > > >> The question: > > ... I have done much research in the field of computer /hardware/ > suitable for commercial Digital Content Creation (P4 Xeon; Wildcat > graphics; ART's "PURE" raytracing PCI render board, etc. ...) and now > look to a better understanding of the choices I might make for a > (presumably) Linux-based OS running on (presumably) Intel 32-bit > hardware SUCH AS these two elements:- > > 1. Kernel Patches - pre-empt and low latency; > 2. File System type - EXT3, ReiserFS, SGi's XFS, HFS, JFS....; > > Would you be able to advsie me on any issues I might need to be aware of > and perhaps any firm decisions you think would be good for me to make, > regarding OS choice and configuration? I realise that all pre-built > workstations supplied by Dell, IBM and HP-Compaq come with RedHat, but, > just for the record, I like Mandrake and I am getting into the > WindowMaker desktop :). > > I also develop my website locally using Apache (but hosted on Freeserve > > :( ), and I understand pre-empt &/or low-latency patches are > > counter-effective for servers. However, as long as no real harm can be > expected, my priority is for graphics and sound creation, editing, > compositing, publishing, etc., while fast response for just developing > html pages isn't. > > I would like to believe that, somewhere, I can get a 'standard' Linux > patched for optimal DCC, with suitable FS type available to choose from > during installation. If not, I will have to find and apply patches, but > I know that with FS type any change demands a reformat, which can really > disrupt a working week! > > >> [End question] > > Yours faithfully, > > James Thompson, > Visual Artist & Musician > H.E. Student of Fine Arts -- Sam Vilain, sam@vilain.net Sure there are dishonest men in local government. But there are dishonest men in national government too. RICHARD M NIXON ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 15:58 ` Sam Vilain @ 2003-01-29 19:31 ` Dieter Nützel 0 siblings, 0 replies; 10+ messages in thread From: Dieter Nützel @ 2003-01-29 19:31 UTC (permalink / raw) To: Sam Vilain, james, reiserfs-list Am Mittwoch, 29. Januar 2003 16:58 schrieb Sam Vilain: > Reiserfs is probably not what you want for doing lots of high volume data. > Reiserfs is good with many small files and general purpose use. It is > actually the slowest filesystem for bulk data. To general ;-) > I'm sure some smartass will probably post some benchmark to prove me wrong, > but SGI have a long heritage of making filesystems for exactly what you > want and so if XFS works I'd say use that. It's pumpin'. Of course more > quantitative comparisons can easily be found on the linux-kernel mailing > list. No. View the ext2/3, Reiser, FS mailing lists. XFS was designed for granted video/audio bandwidth in the first place. Latest ReiserFS 3.x patches catches up and Reiser4 will do some/better. > However, the practical difference between the high performance filesystems > you mentioned, or the difference between running pre-empt or not should be > considered marginal compared to other factors You have to split this into two sentences. "High performance filesystems" and low kernel response (low latency and pre-emption). You will have all three together I "swear";-) Everyone who haven't tested a "multi media"/VIS workstation (3D OpenGL/Mesa (DRI) graphics) without pre-emption and low latency should try. > - such as, the load that > your hard disk controller places on the system and the number of physical > disks you have (and how many RPM they run at). Keep in mind that you can > often nearly double the data throughput of a system by doubling the number > of physical disks in it, and using RAID. Yes. Hard or SoftRAID as you like. > Even with the latest UDMA-133, I haven't seen any IDE system actually > perform without bothering the CPU non-trivially (of course YMMV). Using > SCSI disks and controllers will give you a smoother system ride; which is > why 95% of high-end workstations come equipped with SCSI. You can get > 15,000 RPM U320 SCSI disks, which are f*** fast (though loud). This is > largely because the SCSI protocol was designed properly, IDE/ATA is a > hack. Serial-ATA promises to offer SCSI like host efficiency, but I'll > only believe it when I see it. And at the moment the costs are as bad as > SCSI anyway. Very true. But ATA-RAID (SoftRAID) together with LVM/LVM2/EVMS is OK. > Of course, being able to move around large chunks of data quickly extends > to other parts of the system, too. The bigger and faster the system BUS, > the better. Having `researched hardware platforms' you should know this, > of course. As far as Athlon Motherboards go, Tyan are a reputable vendor > who used to produce Sun clone motherboards - they have a really nice dual > capable system with dual U160 SCSI controllers and 66MHz PCI slots - which > must mean that the SCSI controllers run at that speed. It's definitely > worth the ~$500 price tag. They all pale in comparison to R10000+ based > SGI or Sparc 9+ platforms of course. All AMD 768MPX dual motherboards have 2 PCI64/66 slots. Running a MSI K7D Master-L, (aka MS-6501, Rev 1.0, AMD 768MPX), dual Athlon MP 1900+, 1 GB DDR-SDRAM 266 CL2 (2x 512 MB), Radeon 8500 (DRI CVS), 10k U160 system, here. > From a graphic artist's perspective, you're probably better off buying a > new G4 based system and running MacOS X, you know :-). If you already have the software for Mac, yes. But if you will/have to buy it then you can go fine with Linux. See all the good "Hollywood" reports. > Linux isn't exactly `The Platform' for digital content creation. To general. Some apps are missing, but it catches up very fast. Even with low lateny and pre-emption. With Linux 2.6 still around it _is_ predestinated for this stuff > The Mac's CPU and hardware platform are a lot better at moving data around, In short: NO. You have much more hardware alternatives with Linux. Consider the brand new AMD Hammer waiting for launch. > and if you've got a Mac then you can run Mac OS X, or Linux + Mac OS > inside an emulator (which runs FAST!). Yes. But fast? ;-) Most Linux binary apps are made for x86 systems. > As a final note, keep in mind that a filesystem reformat does not mean a > re-install; keep your partitions as small as practical and you can change > them over individually. See above: LVM/LVM2/EVMS [full quote deleted] Regards, Dieter -- Dieter Nützel Graduate Student, Computer Science University of Hamburg Department of Computer Science @home: Dieter.Nuetzel at hamburg.de (replace at with @) ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 20:20 What Filesystem? James Thompson 2003-01-29 15:28 ` Ross Vandegrift 2003-01-29 15:58 ` Sam Vilain @ 2003-01-29 18:16 ` Valdis.Kletnieks 2003-01-29 19:43 ` Dieter Nützel 2 siblings, 1 reply; 10+ messages in thread From: Valdis.Kletnieks @ 2003-01-29 18:16 UTC (permalink / raw) To: james; +Cc: reiserfs-list [-- Attachment #1: Type: text/plain, Size: 1282 bytes --] On Wed, 29 Jan 2003 15:20:26 EST, James Thompson <james@jtmstudio.fsbusiness.co.uk> said: > ... I have done much research in the field of computer /hardware/ > suitable for commercial Digital Content Creation (P4 Xeon; Wildcat > graphics; ART's "PURE" raytracing PCI render board, etc. ...) and now Hmm.. so we're looking at high-end rendering, which is usually a CPU hog. > 1. Kernel Patches - pre-empt and low latency; so I'm not clear on why you're worried about low latency? Remember that it *does* come with an overhead - the low-latency stuff is good if you're more concerned about fast response than total system load (for instance, on my laptop I'm willing to give up 5% of the CPU if it makes the X server run perceivably faster. If I was doing a lot of rendering, I'd want that 5% for user cycles. Could you be more specific regarding what sort of content you are making? (i.e. single frame images suitable for monitor display (1600x1200 and smaller), or large-format for high-resolution printing (posters, etc), or video, etc..) The resources needed to produce a 10-minute video clip are different from the things you'll need to produce a 4 foot x 5 foot poster at 600DPI. -- Valdis Kletnieks Computer Systems Senior Engineer Virginia Tech [-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --] ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 18:16 ` Valdis.Kletnieks @ 2003-01-29 19:43 ` Dieter Nützel 2003-01-29 20:09 ` Valdis.Kletnieks 0 siblings, 1 reply; 10+ messages in thread From: Dieter Nützel @ 2003-01-29 19:43 UTC (permalink / raw) To: Valdis.Kletnieks, james; +Cc: reiserfs-list Am Mittwoch, 29. Januar 2003 19:16 schrieb Valdis.Kletnieks@vt.edu: > On Wed, 29 Jan 2003 15:20:26 EST, James Thompson <james@jtmstudio.fsbusiness.co.uk> said: > > ... I have done much research in the field of computer /hardware/ > > suitable for commercial Digital Content Creation (P4 Xeon; Wildcat > > graphics; ART's "PURE" raytracing PCI render board, etc. ...) and now > > Hmm.. so we're looking at high-end rendering, which is usually a CPU hog. > > > 1. Kernel Patches - pre-empt and low latency; > > so I'm not clear on why you're worried about low latency? Remember that > it *does* come with an overhead What? --- Sorry, quantify it. I can't. All low latency and pre-emption tests have showed _improved_ throughput (Yes) and much better "multi media" (video/audio) "experience" ;-) No measurable overhead an single and SMP systems (both Athlon). That's why it is in 2.5/2.6. > - the low-latency stuff is good if you're > more concerned about fast response than total system load (for instance, > on my laptop I'm willing to give up 5% of the CPU if it makes the X server > run perceivably faster. If I was doing a lot of rendering, I'd want that > 5% for user cycles. But you want to have the much better "task switching behavior" together with the "brand new" O(1) scheduler. > Could you be more specific regarding what sort of content you are making? > (i.e. single frame images suitable for monitor display (1600x1200 and > smaller), or large-format for high-resolution printing (posters, etc), or > video, etc..) The resources needed to produce a 10-minute video clip are > different from the things you'll need to produce a 4 foot x 5 foot poster > at 600DPI. You mean single vs 2-/4-/8-/etc. (NUMA) SMP systems or even clusters? ;-) Greetings, Dieter -- Dieter Nützel Graduate Student, Computer Science University of Hamburg Department of Computer Science @home: Dieter.Nuetzel at hamburg.de (replace at with @) ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 19:43 ` Dieter Nützel @ 2003-01-29 20:09 ` Valdis.Kletnieks 2003-01-29 20:16 ` Dieter Nützel 0 siblings, 1 reply; 10+ messages in thread From: Valdis.Kletnieks @ 2003-01-29 20:09 UTC (permalink / raw) To: Dieter Nützel; +Cc: james, reiserfs-list [-- Attachment #1: Type: text/plain, Size: 1295 bytes --] On Wed, 29 Jan 2003 20:43:38 +0100, Dieter =?iso-8859-1?q?N=FCtzel?= said: > All low latency and pre-emption tests have showed _improved_ throughput (Yes) > and much better "multi media" (video/audio) "experience" ;-) > No measurable overhead an single and SMP systems (both Athlon). > That's why it is in 2.5/2.6. Yes, but my *point* was that if the box is busy doing a 5-hour rendering job, the low-latency *wont make a difference*. If you're doing audio work or video captures, yes the low-latency/preempt stuff will help. If you're in a single thread that's CPU-bound, there's no need to go installing 2.5 to get it. > But you want to have the much better "task switching behavior" together with > the "brand new" O(1) scheduler. You want that *if* the actual load will benefit from it. See my comment above about rendering... > You mean single vs 2-/4-/8-/etc. (NUMA) SMP systems or even clusters? ;-) Some things are doable with a single CPU, some things need the NUMA stuff, other things can use a Beowulf. Thus the need to quantify what the actual requirements are. Video editing will bottleneck on different things than photorealistic rendering of graphics. -- Valdis Kletnieks Computer Systems Senior Engineer Virginia Tech [-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --] ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: What Filesystem? 2003-01-29 20:09 ` Valdis.Kletnieks @ 2003-01-29 20:16 ` Dieter Nützel 0 siblings, 0 replies; 10+ messages in thread From: Dieter Nützel @ 2003-01-29 20:16 UTC (permalink / raw) To: Valdis.Kletnieks; +Cc: james, reiserfs-list Am Mittwoch, 29. Januar 2003 21:09 schrieb Valdis.Kletnieks@vt.edu: > On Wed, 29 Jan 2003 20:43:38 +0100, Dieter =?iso-8859-1?q?N=FCtzel?= said: > > All low latency and pre-emption tests have showed _improved_ throughput > > (Yes) > > > > and much better "multi media" (video/audio) "experience" ;-) > > No measurable overhead an single and SMP systems (both Athlon). > > That's why it is in 2.5/2.6. > > Yes, but my *point* was that if the box is busy doing a 5-hour rendering > job, the low-latency *wont make a difference*. If you're doing audio work > or video captures, yes the low-latency/preempt stuff will help. If you're > in a single thread that's CPU-bound, there's no need to go installing 2.5 > to get it. > > > But you want to have the much better "task switching behavior" together > > with the "brand new" O(1) scheduler. > > You want that *if* the actual load will benefit from it. See my comment > above about rendering... I am all with you. > > You mean single vs 2-/4-/8-/etc. (NUMA) SMP systems or even clusters? ;-) > > Some things are doable with a single CPU, some things need the NUMA stuff, > other things can use a Beowulf. Thus the need to quantify what the actual > requirements are. Video editing will bottleneck on different things than > photorealistic rendering of graphics. Very true. Regards, Dieter -- Dieter Nützel Graduate Student, Computer Science University of Hamburg Department of Computer Science @home: Dieter.Nuetzel at hamburg.de (replace at with @) ^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2003-01-29 20:20 UTC | newest] Thread overview: 10+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2003-01-29 20:20 What Filesystem? James Thompson 2003-01-29 15:28 ` Ross Vandegrift 2003-01-29 15:32 ` Anders Widman 2003-01-29 19:02 ` Dieter Nützel 2003-01-29 15:58 ` Sam Vilain 2003-01-29 19:31 ` Dieter Nützel 2003-01-29 18:16 ` Valdis.Kletnieks 2003-01-29 19:43 ` Dieter Nützel 2003-01-29 20:09 ` Valdis.Kletnieks 2003-01-29 20:16 ` Dieter Nützel
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.