Home > General Development > IBM’s JFS 1.1.3 Released IBM’s JFS 1.1.3 Released Eugenia Loli 2003-09-06 General Development 23 Comments Dave Kleikamp announced the release of JFS 1.1.3 on the lkml yesterday. Anyone using JFS who recently upgraded to the 2.6-test Linux kernel will want to take note of the new version of the utilities. About The Author Eugenia Loli Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker. Follow me on Twitter @EugeniaLoli 23 Comments 2003-09-06 5:51 pm and if you use it have to pay to sco 650$ Back to the subject , Anyone here is using it , Some bechmarks ? 2003-09-06 6:41 pm Not sure why though. I guess I am not alone because I see very few people who talk about this FS. I use XFS and Reiser all the time, even EXT3 sometimes but for some reason I always gloss this one over. Why is everyone ignoring JFS? Was there a bad review some time ago this is in the back of our heads? 2003-09-06 6:51 pm Go ahead and try it. once and if sco wins, then you can always go back to BSD.by the way sco will run out of microsoft money before that final day in court. 2003-09-06 7:29 pm I am using IBM JFS on 2 IBM Thinkpad A22P with Gentoo and on one desktop computer (using Gentoo as well). On the server I am using SGI XFS. And when I was installing my Thinkpad to serve as an reference (for quicker install of Gentoo) for the other 3 Thinkpad’s floating around in my company, I was looking at the Thinkpad and decided to use an IBM filesystem on an IBM notebook. One of my employees wanted to use ReiserFS because she heard the speed of ReiserFS is great. But after serval crashes (once total powerloss in the complete building and the battery went offline as well) and serval –rebuild-tree with the endresult: everything was in “lost+found” with random names, she gived IBM JFS an try and since that time we both are using IBM JFS on our Thinkpads. It is not that fast as ReiserFS or SGI XFS, but it is dam reliable! We once did an fsck on an mounted root filesystem and it fixed the problem (fsck warned us, that it could be dangerous…). I have very bad experiance with ReiserFS on an mobile computer. SGI XFS is great, but can lead into big trouble if you loose power and data is still not flushed to disk. Anyway… IBM JFS is an very nice fs and beside SGI XFS and EXT3, it’s the only fs (on Linux) I trust cheers SteveB 2003-09-06 7:57 pm IBM keeps slapping SCO allover the face. Come to think of it, I’ve never used a JFS file system. I should give it a try someday. Regards, Mystilleef 2003-09-06 8:19 pm If you choose to use our (SCO’s) IP, you will do so at your own risk and with the distinct possibility of being sued by us. SCO will tolerate IBM releasing our IP to open source. On the other hand, JFS is pretty cool. I use it regularly. 2003-09-06 8:23 pm I was reading over some recent JFS benchmarks; Aparently, when you run JFS and Kernel 2.6 you get hella good results. I’ve been using JFS on my AIX boxes for years (I know I am so 1337). Anyway, now may be the time for JFS on Linux. 2003-09-06 9:42 pm Why is everything about benchmarks ? Doesn’t reliability count ? JFS is used much in the telecom area just because of that. As many know, telecom applications have some obscene requirements for uptime/production time, and JFS actually shines here, it’s very reliable, and does some real hard work ensuring the integrity of the data. 2003-09-06 9:53 pm they are all so horny about benchmarks, because they use their computer as an toy or to show the world how 1337 they are! if they would use their computer for productive environments, then they would put reliability bevore speed. 2003-09-06 11:40 pm Or maybe because desktop users don’t have the same requirements…? I don’t think I would use ReiserFS on a server, but I sure do on my desktop computer except for my home partition where my data is more important than speed. 2003-09-07 3:46 am How can SCO claim that JFS is their intellectual property when the Linux version is based off the OS/2 eServer which in turn is a clean implementation of JFS. Regards to BSD, IIRC, there were only three files that infringed on the SYSV license, the settlement was done on the basis that the BSD community would replace the 3 files. These 3 files have been replaced, hence, SCO has no case what so ever against any of the *BSDs. 2003-09-07 4:33 am // Regards to BSD, IIRC, there were only three files that infringed on the SYSV license, the settlement was done on the basis that the BSD community would replace the 3 files. These 3 files have been replaced, hence, SCO has no case what so ever against any of the *BSDs. // LOTS of people have been saying this, and its completly wrong. ALL of UNIX belongs to SCO (so they say), that includes MODERN unix. That BSD lawsuit was a long time ago. So at the date of the lawsuit, BSD didnt infringe on UNIX copyrights, but time has passed, and code has been added, and that code could (according to SCO) be infringing. The only way BSD would be 100% clear is if NOT A SINGLE LINE OF CODE was added since the AT&T lawsuit. That said, SCO is full of it. 2003-09-07 5:46 am LOTS of people have been saying this, and its completly wrong. ALL of UNIX belongs to SCO (so they say), that includes MODERN unix. That BSD lawsuit was a long time ago. So at the date of the lawsuit, BSD didnt infringe on UNIX copyrights, but time has passed, and code has been added, and that code could (according to SCO) be infringing. The only way BSD would be 100% clear is if NOT A SINGLE LINE OF CODE was added since the AT&T lawsuit. That said, SCO is full of it. SCO only owns the code to the AT&T line, OpenGroup owns the UNIX trademarks and specification (aka UNIX 98/UNIX 95). “The first CDROM (and general net-wide) distribution was FreeBSD 1.0, released in December of 1993. This was based on the 4.3 BSD Lite (“Net/2”) tape from U.C. Berkeley, with many components also provided by 386BSD and the Free Software Foundation. It was a fairly reasonable success for a first offering, and we followed it with the highly successful FreeBSD 1.1 release in May of 1994. Around this time, some rather unexpected storm clouds formed on our horizon as Novell and U.C. Berkeley settled their long-running lawsuit over the legal status of the Berkeley Net/2 tape. A condition of that settlement was U.C. Berkeley’s concession that large parts of Net/2 were “encumbered” code and the property of Novell, who had in turn acquired it from AT&T some time previously. What Berkeley got in return was Novell’s “blessing” that the 4.4 Lite release, when it was finally released, would be declared unencumbered and all existing Net/2 users would be strongly encouraged to switch. This included us, and we were given until the end of July 1994 to stop shipping our own Net/2 based product. Under the terms of that agreement, we were allowed one last release before the deadline and that became FreeBSD 188.8.131.52, the culmination of our year’s work with Net/2 and generally considered by many to be a significant project milestone for stability and general performance.. We then set about the arduous task of literally re-inventing ourselves with a completely new and rather incomplete set of 4.4 Lite bits. The “Lite” releases were light in part because Berkeley’s CSRG had removed large chunks of code required for actually constructing a bootable running system (due to various legal requirements) and the fact that the Intel port of 4.4 was highly incomplete. It took us until December of 1994 to make this transition, and in January of 1995 we released FreeBSD 2.0 to the net and on CDROM. Despite being still more than a little rough around the edges, the release was a significant success and has since been followed by the more robust and easier to install FreeBSD 2.0.5 release in June of 1995.” So, unless you’re running a 1.x version of FreeBSD (which is highly unlikely), you don’t owe SCO a dime. SCO is running out of money fast and instead if improving their product and making their case to customers to why they (customers) should buy their (SCO) over the competition, they’ve decided to take the easy route, grab a bunch of half-baked arguments, string them together, get the marketing machine working overtime and just hope that the victim, in this case IBM, will simply crumble. Using SCO’s logic, If I create a driver for UNXWare, it instantly becomes the property of SCO because according to their logic, the driver is derived from SCO “technology”. 2003-09-07 7:40 am I think the point the original poster is making is that history has nothing to do with it. If code from SCO UNIX has made its way into the *BSDs since the lawsuit, the results of the lawsuit do not protect the BSDs. Its a fresh infringement case. For example, if a piece of the SMP code that SCO alleges turned Linux from a “mom and pop OS” into an enterprise OS also found its way into SMPng, then there is a problem. That said, SCO is full of it. Read the Wired Interview with Darl McBridge. Not only does the interviewer deal with DM in an rather harsh manner (“So, how does it feel to be remembered as the guy who killed Open Source?”) but DM’s answers are totally contradictory to what SCO is saying today. DM, in that interview, says that the lawsuit won’t kill Linux or the GPL, but make it stronger because it will ensure there is no infringing code. However, now SCO is saying that the GPL is invalid! Its all total horse manure… 2003-09-07 1:48 pm Hi Just wanted to know why XFS is not as reliable as JFS as both of them are journaled? Thanks. 2003-09-07 2:22 pm I think the point the original poster is making is that history has nothing to do with it. If code from SCO UNIX has made its way into the *BSDs since the lawsuit, the results of the lawsuit do not protect the BSDs. Its a fresh infringement case. For example, if a piece of the SMP code that SCO alleges turned Linux from a “mom and pop OS” into an enterprise OS also found its way into SMPng, then there is a problem. And even if it were, then it would violate the GPL license that the SMP code was licensed under. Either way, BSD would NEVER be affected. That said, SCO is full of it. Read the Wired Interview with Darl McBridge. Not only does the interviewer deal with DM in an rather harsh manner (“So, how does it feel to be remembered as the guy who killed Open Source?”) but DM’s answers are totally contradictory to what SCO is saying today. DM got what he deserved and it is about time the media did what Wired did during the interview and actually say the hard statements and not hold back. I know it is pretty normal in the US media to suck up to people in “positions of authority”, but this is now a new era of reporting, what they write can’t simply be pushed aside as “for domestic consumption”. DM neds to be held accountable for what he says, whether the interviewer be from the US or some where else. DM, in that interview, says that the lawsuit won’t kill Linux or the GPL, but make it stronger because it will ensure there is no infringing code. However, now SCO is saying that the GPL is invalid! Its all total horse manure… DM is a moron, plain and simple. Maybe he should head off to the bible and read the ten commandments because he EASILY breaks, “They shall not bare false witness”. It seems that he is all high and holy but when it comes to bringing values and ethics from his faith, they’re put right out to pasture right next to the ability to see the bloody obvious, namely, SCO has no case and is a dying company running on empty. 2003-09-07 3:06 pm I did some benchmarking with the 2.4.18 kernel on Gentoo 1.3 some time ago, comparing XFS, ReiserFS, Ext3 and JFS speeds. Just a Python script doing some file access: creating files, moving them around, reading them, searching for them. Nothing overly scientific, though. JFS won hands down. Unfortunately, the version of JFS that came with that version of Gentoo had a horrible glitch that occasionally had it core-dumped and freeze up from time to time. The JFS journaling would bring the file system back to normal on reboot, though I lost a few files here and there. It was, as I recall, a known bug that had been fixed in later releases of JFS, but not in the version included with Gentoo. That made me switch to Ext3 just to get things quickly up and working again, but I would gladly run JFS over ReiserFS or XFS. JFS is not a “legacy” product. It’s more than able to compete with the more popular file systems. 2003-09-07 4:06 pm Ok then i’ll try to format a new drive with it . (Seagate baracuda) It could be nice from RedHat to let you choose the FS you want (like they do on Mandrake,Slackware). I know that you can put a parameter when install it but is not visible. From my experience ext3 is really solid , I turn off the computer (rh8.0) without shut down at home (every day) and there are no files missing . Also tried MDK8.0 with reiserfs and did the same procedure(turn off without shutdown) but sometimes the files disapeared or were courupted . I think they fixed in the new versions (didn’t tried yet). Ok now want to try something new(for me) and solid JFS sounds good for this 2003-09-07 6:37 pm Is there a particular reason why you turn your computer off without shutting it down? IMHO it’s not so sharp, but you can do whatever you like! 2003-09-07 7:21 pm <quote>Regards to BSD, IIRC, there were only three files that infringed on the SYSV license, the settlement was done on the basis that the BSD community would replace the 3 files. These 3 files have been replaced, hence, SCO has no case what so ever against any of the *BSDs. </quote> People must have gotten tired of reading the articles. In an interview a SCO rep (can’t remember who) claimed that the BSD folks haven’t lived up to the agreement made with AT&T back during the last trial. SCO is coming for you too guys… I should put in an obligatory reference to M$ (I know much that spelling ticks off the idiots who read these pages) 2003-09-07 8:40 pm This is how it should work a journaled system , imagine that that system isn’t a home pc but a critical server Take the worst scenario :if the power fails (ups fails too) and your server is turned off (without proper shut down) then the fs should be in a consistent state . So is a good test i think 2003-09-08 12:35 am My experiences with JFS have been horrible. I’m always on the last stable release, but it has never ever actually worked for me. When my PC crashes, on reboot, JFS says all is fine. After working for a few minutes with the crashed file system, the kernel locks up with a JFS error. So what I do when I’ve had a crash, is boot, unmount the JFS file systems, manually force check them, and after that they’re fine. I also have lost files some times or files have been corrupted. I’ve not used ext3 or any of the other logging file systems, so I don’t know if JFS is better or worse, but if I could, I would switch. I don’t have the HD space, nor time, to move over big file systems unfortunately. I’ve used NT 3.51 until Windows 2000 and I can’t remember a single problem with NTFS (same PC, before I switched 100% to Linux on that PC). Never lost a file, never had to unmount and manually check a file system there. 2003-09-08 9:33 am I hadn’t tried it in what seems like ages, but it was more then a few releases of Mandrake ago – maybe two plus years? It was much newer then, wasn’t yet in mainline 2.4, and I think Mandrake just started offering it. I tried it out, and ugh – performance sucked and after switching off power on purpose I lost some files, nothing overly major – but I wasn’t impressed. JFS is, on AIX and OS/2 I’d imagine, mature and robust – its nicely aged, so I’m more then willing to give it the benefit of a doubt. My experience was from very early code, and based on the comments thus far I’d suspect alls much better now. Someone earlier claimed JFS is fast, contrary to most benchmarks – was it the latest release? Someone else mentioned stability vs performance, I agree that between the two stability for servers is paramount.