Slashdot Mirror
Reiser4 Benchmarks
unmadindu writes "Hans Reiser has benchmarked Reiser4 against ext3 and Reiserfs 3. Reiser4 turns out to be way faster than V3, and for ext3, why don't you check out the results yourself ? Hans Reiser states, "these benchmarks mean to me that our performance is now good enough to ship V4 to users", and he will be probably sending in a patch within the next couple of weeks to be included in the 2.6/2.5 kernel."
My one concern is reliability and recovery from failure; I've had a few cases where my belief in ReiserFS has been questioned; however I can't get Ext3 to build on larger than 500GB arrays.
At this point I'd happily choose based on reliability/recoverability/stability not raw speed.
Does anyone know if there will be a conversion utility available - i.e, to convert ReiserFS v3 partitions to v4?
This space intentionally left blank.
hey, I can live with an unstable gnome or Kicq, but a beta filesystem?? no thanks dude!
After reiser4, what filesystems are actually decent competition for it? It'd be nice for OSS to claim not only the best web server (apache), best kernel, and best filesystem.
Slashdot: Where people pretend to be twice as smart as they really are by behaving like children.
I am curious as to whether there are any projects to port Reiser4 to *BSD, particularly FreeBSD 5.x. Does anyone have any thoughts on how difficult a port might be? Can somone more versed in filesystems on *nix enlighten me as to the implimentation differences?
**AA: a bunch of mindless jerks who'll be the first against the wall when the revolution comes
In case anybody cares, "strelka" means "arrow", and "belca" means "squirrel"
Wonder what naming system they're using. I use names from Alice in Wonderland.
The statistics on that page are measured in seconds, no? So larger numbers are worse.
The comparisons are done with [foreign filesystem] divided by reiser4.
One would think that numbers greater than one, where the foreign filesystem has a long running time and reiser4 a short one, would be the ones that benefit reiser4.
Yet the numbers *less* than one are green, where Hans says reiser4 is considered better.
What's going on?
(Incidently, after having a friend lose a filesystem to buggy reiser code, I'm a bit inclined to wait until people have *seriously* hammered on this).
May we never see th
While great, this announcement/benchmark/statement does not mean that ReiserFS V4 is ready for production use, just that it is fast. It needs a lot more bug testing before then, so don't rush out and mass-convert to V4 just yet! See here for the full thread, rather than just the first post...
* Several monkeys are here, playing banjos and wearing small hats.
I realise that it is a bit early to adopt V4, but stable issues aside, which filesystem would YOU choose to for database volumes for fx. Oracle or MySQL?
my sig
How does it compare against everyone's favorite, XFS?
Prescriptive grammar:linguistics
So he's submitting it to 2.6, but what are the chances it'll get submitted? Isn't this what caused all of Reiser's bitching a couple of years ago? He waited to long to get RFS into the kernel and ran into the feature freeze, and then pitched a hissy fit.
.
I'm not sure about the exact reasons why they don't support various other filesystems. The default bootup sequence of a RH system uses an initial ramdisk, and actually scans each partition available to find out where they should be mounted (they created nash, NAno SHell, which is just simple support for shell commands as well as fs label scanning). That's why you see the LABEL=/ in your /etc/fstab on a RH system. ResiserFS didn't support filesystem labels until 3.6, so using this setup could mess things up (with 3.5 or older), and justifies your point about having to "jump through hoops" to get reiserfs working. The simplest way I found to move to reiserfs was to change all the LABEL=??? specifications to actual device files, boot from a recovery disk, move everything around while reformatting the partitions as another filesystem, then finally rebooting.
Anybody know what, if any, features are being added for the laptop user? Last time a checked, journaled filesystems, like ext3, were generally a no-no if you wanted you battery to last.
Maybe a filesystem just for laptop/tablet pc users?
I attended Hans' presentation at Linuxtag.
Basically, reiser4 is optimized for the case where you unpack a large tarball, say the Linux kernel, and have enough memory to hold it all in cache, which is true for most of us these days. reiser4 will then choose the optimal disk layout for these files and flush them to disk.
Hans also has aspects of a log structured file system in reiser4, which means you don't write to the file, you write to a log file which basically encompasses the whole disk. The up side is that you mostly write linearly, the down side is that the files get badly fragmented if they are updated at all. Most files are not updated, just written once at installation of the package. The files that are updated frequently tend to be source code from CVS, which are small enough to fit in memory completely and have reiser4 choose an optimal disk layout again.
The case where this model completely sucks is the case where you update many portions of a large file. For example, running an SQL database with files on a reiser4 file system as backend, or maybe a DNS server with DDNS, or a berkeley db backend for Postfix or qmail to keep the SMTP AUTH users or something. Also, log files will probably be badly fragmented.
Hans proposes to have something like a transparent defragmenter running in the background, which he calls "repacker". This would run in the kernel space, as part of the file system, and defragment badly fragmented files that are accessed frequently. This would solve most of the down sides of his approach, but this repacker is not finished yet.
My personal view of reiser4 is: it looks like it is optimized to perform well in benchmarks. It tries to be fastest for updating databases, but buys the performance by being slower when reading the data afterwards. The critical question is whether the repacker can alleviate these concerns, and as long as it is not finished, reiser4 is basically out of the question except for a little testing here and there. I reckon reiser4 would be a great filesystem for keeping your mozilla and gcc CVS checkout handy. But until the repacker is done, I will not even use it for testing, because the repacker really is the crucial component that makes or breaks this.
By the way: my previous experiences with reiserfs were less than stellar. Some people call it shredderfs instead. The main complaint with reiserfs is and always was that the fsck is not nearly as trustworthy or stable as the one from ext2/ext3. So even if I use reiserfs at all, it's only for data I can afford to lose completely, like my CVS checkouts or the squid cache directories or something like that.
The benchmarks do look good though, and I am glad that at least someone is still trying major innovations in this area. Since most Unix vendors or divisions are no longer profit centers, file system innovations have largely stalled or moved to specialized companies who regard them as proprietary (Veritas) instead of releasing them as free software like IBM and SGI did.
It's still not time to swap it for ext3 for general use.
The first table with the mixed file sizes is the most compelling. The fact that reiser4's Create and Copy times are less than a third of the ext3s in real_time is impressive.
But the fact that the CPU consumption on Read is double that for R4 as it is for ext3 is a serious problem. On a 1.3 Ghz machine saturating a generic UDMA 100 60G bus on RH 9.0 it's about 10% of the CPU, so the home user might not care. For a system capable of delivering serious data (like a 4 drive, 15k rpm SCSI RAID array @~3 times the read throughput) going from 30% CPU to 60% CPU usage is a definite problem. Even with a 2.6 Ghz cpu it would still move from chewing up 15% to 30%. I know these numbers don't scale exactly, but they could in fact scale ugly depending on how much CPU is dedicated to communicating with the hardware and how much is in fiddling with the filesystem. My production boxes spend > 80% of their disk activity reading, so I'm not yet inspired to go out and spend the time running benchmarks on highperf. systems just yet.
Nevertheless, I always admire it when a new version of software comes out and it's noticeably faster than the old
Apple benchmarked their new G6 processor against the latest 10 GHz Pentium V. They say that despite its lower clock speed, it runs their suite of PhotoShop 8 filters almost four time faster than the Pentium.
Seriously, Hans Reiser is benchmarking his own file system, and he's using benchmarks that make his system look good. Like the SpriteLFS, his filesystem has a log structure for sequential writing, which makes it look really good in tests like he performed where you write the files once.
Compare a database load, where you write small chunks of big files all the time. Without the repacker (like the cleaner in LFS), the disk becomes horribly fragmented. With the repacker, you have to include the slowdown of this background process defragging your hard disk. Ick.
I'll trust his benchmarks when he presents a final, stable release, with the repacker on, and tests it under workloads such as would be encountered on a server. I might use it on my homebox even if it sucks on a server, but it would be nice to know that he considers his structure's impact on other workloads.
I hereby place the above post in the public domain.
That was V3. V4 is an atomic filesystem, which means that every filesystem operation is performed as a fully atomic transaction. This is more secure than the guarantees of data journaling, as data journaling doesn't necessarily guarantee that the write will complete.
The reason we are able to do this in V4 but not in V3 is that V4 uses what I call wandering logs. With wandering logs, instead of copying data first to the journal, and then after commit copying it from the journal to the rest of the filesystem, (thereby writing the data twice), we just change our definition of where the journal is. I don't think that data journaling is worth going half-speed for most users. With V4, we not only don't go half-speed, we go faster than V3 ever went.
For more details, please take a look at http:www.namesys.com/v4/v4.html
It is important that you use a distro that bases their kernel on 2.4.18, or later, when using ReiserFS. There exists a distro that bases their enterprise server kernel on 2.4.9, and intentionally declines to add any reiserfs bugfixes since then. (This is the same distro that once shipped their kernel with the ReiserFS debugging code turned on so that we would go slow.) Do NOT use their kernel with ReiserFS. Generally when someone reports that they are having really bad experiences with ReiserFS, it turns out they are using that kernel.
I generally recommend using the latest official kernel from Marcelo, and not any distro kernels, but the SuSE kernels tend to have effective ReiserFS support also, and not everyone out there shares my non-technical preference for a common community developed kernel.
Root is a prime candidate for a small (100MB should do it) ext2 system mounted "sync". You don't need decent write performance on root; in fact, many sysadmins make it read-only. Journalling is pointless if you're only writing to the filesystem once every 6 months to add a new user.
On the contrary, that's exactly the case where you should always journal, and with full data journalling. You don't care about write performance, since you hardly ever do it, but you do care at lot about keeping your root filesystem consistent.
Have you got your LWN subscription yet?
I get reports (not verified by me) that ReiserFS V3 is an order of magnitude faster when used as a backend for an XML database than relational databases that were tried. So, if your data happens to have a hierarchical structure, or, you can put it into one, then you are likely to get a performance gain. If your data does not have a hierarchical structure, then you need to wait for V6 where we plan to expand the semantics.
If you want to be able to "cat filenameX/..owner" to see who owns "filenameX", you need to use V4.