Re: FreeBSD's UFS vs Ext4
Using a phoronix link as an example, ext4 still has some pretty bad data loss bugs: http://www.phoronix.com/scan.php?page=news_itempx=Nzk0OA imo: data security of ufs speed of ext4 ymmv ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On 9 February 2010 16:11, Giorgos Keramidas keram...@ceid.upatras.grwrote: On Mon, 8 Feb 2010 12:58:07 -0700, Chad Perrin per...@apotheon.com wrote: On Mon, Feb 08, 2010 at 03:00:00PM +, Frank Shute wrote: AFAIK, the system compiler is going to be clang in the future and for ports you'll install a compiler from ports. Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. There are no concrete plans to ditch gcc from the base system. The GPLv3 license of GCC and binutils does pose a few interesting problems. On the other hand, Clang is a nice project, whose license *is* compatible with a BSD-style license. Several FreeBSD developers have tried building the base system with it and are active at the development forums of Clang. So it's probably a safe assumption to make that Clang is not ready *yet*, but may be an interesting alternative to GCC in the near future. I believe that CLANG will generate quicker code as well from what I have read about it. Which will be nice ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On 9 February 2010 01:54, J65nko j65...@gmail.com wrote: On Mon, Feb 8, 2010 at 5:46 AM, alex a...@mailinglist.ahhyes.net wrote: I do suspect personally that the ext4 filesystem is the reason for the difference here, since ext4 has a number of features such as deferred disk writes etc. Even deleting a large file off that raid array I can see a difference, prior to reformatting, i deleted a 190GB file off the raid, under UFS the delete took quite some time (well over 10 seconds), under ext4 the deletion of the same size file took about 3 seconds. But what I said with ext4 being faster then the aging UFS still rings true in my mind, look at the recent Phoronix benchmarks for yourself and see (10 pages of benchmarks). http://www.phoronix.com/scan.php?page=articleitem=freebsd8_benchmarksnum=1 (skip to page 7 of the benchmarks if you want to see the I/O stuff relating to disk performance) According to the first page they used the default configuration of all benchmarked OS'es. And what is the default mount option on Linux async The FreeBSD man page for mount describes this async option as follows: async All I/O to the file system should be done asynchronously. This is a dangerous flag to set, since it does not guar- antee that the file system structure on the disk will remain consistent. For this reason, the async flag should be used sparingly, and only when some data recov- ery mechanism is present. The OpenBSD man page has the following additional remark: The most common use of this flag is to speed up restore(8) where it can give a factor of two speed in- crease. Conclusion: you cannot compare filesystem performance, when you give one a unfair speed advantage of what could be a factor two. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org you are of course entirely correct, however one of the goals of more modern file systems eg ext4 is to make async safe to use, because of this speed up. At the end of the day faster is faster simple as. Having said that it would be nice to see a gjournaled ufs system for comparison, as well as zfs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
For what is worth these are the results on my Lenovo Thinkpad T500 with zfs. http://global.phoronix-test-suite.com/?k=profileu=thuglife-5875-16786-4629 dmesg | grep ada0 ada0 at ahcich0 bus 0 scbus0 target 0 lun 0 ada0: WDC WD2500BEKT-00A25T0 01.01A01 ATA-8 SATA 2.x device ada0: 300.000MB/s transfers (SATA 2.x, UDMA6, PIO size 8192bytes) ada0: Command Queueing enabled ada0: 238475MB (488397168 512 byte sectors: 16H 63S/T 16383C) zfs prefetch off zfs checksum on | fletcher4 zfs compression on | lzjb vfs.zfs.arc_min=64M vfs.zfs.arc_max=512M stock ufs FBSD http://www.phoronix.com/scan.php?page=articleitem=linux_bsd_opensolarisnum=6 Apples and oranges, I know, the point is I don’t feel that the IO performance is lagging on my laptop. On Tue, Feb 9, 2010 at 9:33 AM, krad kra...@googlemail.com wrote: On 9 February 2010 01:54, J65nko j65...@gmail.com wrote: On Mon, Feb 8, 2010 at 5:46 AM, alex a...@mailinglist.ahhyes.net wrote: I do suspect personally that the ext4 filesystem is the reason for the difference here, since ext4 has a number of features such as deferred disk writes etc. Even deleting a large file off that raid array I can see a difference, prior to reformatting, i deleted a 190GB file off the raid, under UFS the delete took quite some time (well over 10 seconds), under ext4 the deletion of the same size file took about 3 seconds. But what I said with ext4 being faster then the aging UFS still rings true in my mind, look at the recent Phoronix benchmarks for yourself and see (10 pages of benchmarks). http://www.phoronix.com/scan.php?page=articleitem=freebsd8_benchmarksnum=1 (skip to page 7 of the benchmarks if you want to see the I/O stuff relating to disk performance) According to the first page they used the default configuration of all benchmarked OS'es. And what is the default mount option on Linux async The FreeBSD man page for mount describes this async option as follows: async All I/O to the file system should be done asynchronously. This is a dangerous flag to set, since it does not guar- antee that the file system structure on the disk will remain consistent. For this reason, the async flag should be used sparingly, and only when some data recov- ery mechanism is present. The OpenBSD man page has the following additional remark: The most common use of this flag is to speed up restore(8) where it can give a factor of two speed in- crease. Conclusion: you cannot compare filesystem performance, when you give one a unfair speed advantage of what could be a factor two. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org you are of course entirely correct, however one of the goals of more modern file systems eg ext4 is to make async safe to use, because of this speed up. At the end of the day faster is faster simple as. Having said that it would be nice to see a gjournaled ufs system for comparison, as well as zfs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, 8 Feb 2010 12:58:07 -0700, Chad Perrin per...@apotheon.com wrote: On Mon, Feb 08, 2010 at 03:00:00PM +, Frank Shute wrote: AFAIK, the system compiler is going to be clang in the future and for ports you'll install a compiler from ports. Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. There are no concrete plans to ditch gcc from the base system. The GPLv3 license of GCC and binutils does pose a few interesting problems. On the other hand, Clang is a nice project, whose license *is* compatible with a BSD-style license. Several FreeBSD developers have tried building the base system with it and are active at the development forums of Clang. So it's probably a safe assumption to make that Clang is not ready *yet*, but may be an interesting alternative to GCC in the near future. pgp1WqULYU9TE.pgp Description: PGP signature
Re: FreeBSD's UFS vs Ext4
On Monday 08 February 2010 05:46:07 alex wrote: Pieter de Goeje wrote: The fact that the limit is 86MB/sec (which is very low for a raid0 array) makes me think the box suffers from sub optimal network performance during a simple stream test like yours. This could be due to FreeBSD having a poor network driver for your particular NIC or could be due to insufficient tuning of the TCP parameters for this particular test. Hi Pieter. You are right about there being a number of possibilities, however: *The same machine, which over the years has had a number of revisions of freebsd on it (have buildworlded the thing from 7- 7.1 - 7.2 - 8), the performance was always roughly the same amongst the versions, I dont agree with the possibility of the ftp server being 'slow' as I am the only person who copies data to that machine, and the machine is always under a very low (almost non existent) load. That's no reason to rule out inefficient design in the FTP server. I could write a program that sends 1b/sec over the network which serves one user and uses no cpu. * Network card is an Intel Pro 1000, on the server. This is a PCI card (not pci-e), so I believe PCI bus bandwidth limitations may be responsible for me not being able to achieve the maximum 100MB/s network rate (as you mention that 86MB/s is slow for raid0) * The intel network card driver on freebsd and linux are both fairly rock solid and well written. I dont see it being an issue with NIC drivers (they are not vastly different). Solid does not mean high performance. For example, to get maximum performance out of my em nics I have to increase the number of tx and rx descriptors from 256 (default on FreeBSD) to 4096. * Both OS's were stock standard installs, no jumbo frames enabled, no fiddling with sysctl network values. So you haven't tried to improve the performance. Nor have you tried to find out why performance is sub optimal. I am happy with 86MB/s anyway, It's a huge improvement of the 60MB/s barrier I could never get past when that machine was running FreeBSD. To get the rest of the speed, I'd probably have to install a pci-e card on the server. I do suspect personally that the ext4 filesystem is the reason for the difference here, since ext4 has a number of features such as deferred disk writes etc. Ok, what did you actually test? File upload? Download? How big was the testfile? Was the file in cache?. Even deleting a large file off that raid array I can see a difference, prior to reformatting, i deleted a 190GB file off the raid, under UFS the delete took quite some time (well over 10 seconds), under ext4 the deletion of the same size file took about 3 seconds. File deletion speed is relevant how? But what I said with ext4 being faster then the aging UFS still rings true in my mind, look at the recent Phoronix benchmarks for yourself and see (10 pages of benchmarks). http://www.phoronix.com/scan.php?page=articleitem=freebsd8_benchmarksnum= 1 (skip to page 7 of the benchmarks if you want to see the I/O stuff relating to disk performance) The phoronix disk benchmarks are of little value in this case. I'll be happy to explain why if you're interested. Further discussion is useless unless you go back and redo the tests, this time trying to isolate the cause instead of making baseless assumptions about the performance of various FreeBSD subsystems. -- Pieter de Goeje ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 08, 2010 at 10:01:05AM +1100, alex wrote: Frank Shute wrote: On Mon, Feb 08, 2010 at 01:41:29AM +1100, alex wrote: Hi Guys, Today I reformatted a machine (network server) thats run FreeBSD nonstop for at least the last 3 years and installed linux on it. I have a raid 0 setup with 2 hard disks in the very same machine. So you had a machine that had run non-stop for 3 years yet you replace the OS. Clever. Yes I replaced the OS. Because the box was to also be a PBX (running asterisk, instead of just being a file server/web server for running local web apps). I was continually getting coredumps with asterisk. After filing numerous bug reports and hitting dead ends with the asterisk devs, I had enough, because none of them knew how to debug the problem under freebsd, I got fed up and moved the box over to linux, and to my surprise, no more core dumps. Fair enough. I see a number of factors putting freebsd behind: * The teams stubbornness with compiler/base tools (wont move away from gcc 4.2.1 because they just cant accept the GPL2) They don't like the license, that's not stubbornness. Wow thats a good reason to use ancient compilers and assemblers. AFAIK, the system compiler is going to be clang in the future and for ports you'll install a compiler from ports. * The teams stubbornness with the base system binutils (which cause mplayer and other multimedia applications not to build, unless a newer version is installed) Nonsense. You dont see having a set of binutils thats not SSE3 or SSE4 capable as a problem? It's nonsense? I'm not saying that. I don't remember having to install new binutils to install mplayer. Using such an old compiler must have a performance impact on the OS. I say this because compilers improve over time, they generate better, tighter, more optimized code. The binutils shipped with freebsd is more than 5 years old now. A codes age has nothing to do with it's performance. Clearly you know nothing about how compilers generate and optimize code. If this isnt a problem, why would new versions of gcc and binutils continue to surface. Well I can see three obvious reasons, improved code generation, bug fixes, new features. There isn't some vast jump in performance provided by an up to date and buggy as hell version of gcc. The improved code performance isn't worth swallowing some daft, verbiose, impenetrable licence for either. It's not just my personal test that has shown that linux is ahead in numerous areas (performance wise), but the recent phoronix benchmarks that were released when FreeBSD 8 came out, were pretty damning. Link please. Sure, no problem, enjoy: http://www.phoronix.com/scan.php?page=articleitem=freebsd8_benchmarksnum=1 Go on, I am waiting for you to poke holes and attempt to totally invalidate those benchmarks too. On the whole, I don't believe in benchmarks. They don't even tell half the story when it comes to choosing an OS and they're always rather dubious IMO. I'd like to see what the FreeBSD team has to say on this. Alex Despite your FreeBSD T-shirt ownage, your post is a troll. Nobody's interested in your bogus benchmarks opinions on matters that you are not knowledgeable of. Regards, I guess you cant see the difference between a troll and a complaint. I have been using freebsd since the 4.x days. It seems you have quite a chip on your shoulder, frank. Heh, I resemble that remark! I'm well balanced - chip on both shoulders ;) FYI, I've used FreeBSD since 4.3 and before that I used Linux. Linux has a rather nasty thrown together feeling about it in comparison and the scheduler on Linux in those days was bloody useless: it had trouble handling more than one task. So it's swings and roundabouts. Linux maybe more performant but it's got (had) it's problems. Don't know about nowadays but I've got no reason to go back and try it again ATM. If you do, good luck to you! Alex. Regards, -- Frank Contact info: http://www.shute.org.uk/misc/contact.html ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 08, 2010 at 10:01:05AM +1100, alex wrote: Frank Shute wrote: On Mon, Feb 08, 2010 at 01:41:29AM +1100, alex wrote: I see a number of factors putting freebsd behind: * The teams stubbornness with compiler/base tools (wont move away from gcc 4.2.1 because they just cant accept the GPL2) They don't like the license, that's not stubbornness. Wow thats a good reason to use ancient compilers and assemblers. Sometimes, license choice *is* a good reason to make some sacrifices in short-term convenience. -- Chad Perrin [ original content licensed OWL: http://owl.apotheon.org ] pgpYUxr0eiWEh.pgp Description: PGP signature
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 08, 2010 at 03:00:00PM +, Frank Shute wrote: AFAIK, the system compiler is going to be clang in the future and for ports you'll install a compiler from ports. Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. -- Chad Perrin [ original content licensed OWL: http://owl.apotheon.org ] pgpVmwOwjAumb.pgp Description: PGP signature
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 8, 2010 at 12:58 PM, Chad Perrin per...@apotheon.com wrote: Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. A quick search yielded these links: http://wiki.freebsd.org/BuildingFreeBSDWithClang http://www.linuxtoday.com/developer/2009051100335NWBD Cheers, Ben ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Monday 08 February 2010 12:54:26 Pieter de Goeje wrote: Even deleting a large file off that raid array I can see a difference, prior to reformatting, i deleted a 190GB file off the raid, under UFS the delete took quite some time (well over 10 seconds), under ext4 the deletion of the same size file took about 3 seconds. File deletion speed is relevant how? It can be, depending on the workload. I (as a Linux user) moved from ext3 to xfs, ignoring the warnings about file deletion [being slow]. Now I _kind of_ regret it. Seems I have more than one program on my laptop that deletes files (kmail's email-expiration thing comes to mind). I also work on a project that creates large log files an deletes them (periodically). When all these programs meet, I go for a coffee. :) -- Mihai Donțu ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On 2010-02-08, at 2:58 PM, Chad Perrin wrote: Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. Here's last year's status report where they talk about it: http://lists.freebsd.org/pipermail/freebsd-stable/2009-May/049873.html It was also in a presentation at BSDCan last year. -- Regards, Derek Buttineau Internet Systems Developer Compu-SOLVE Internet Services Compu-SOLVE Technologies, Inc Phone: 705-725-1212 x255 E-Mail: de...@csolve.net ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 08, 2010 at 01:11:26PM -0700, Ben Schumacher wrote: On Mon, Feb 8, 2010 at 12:58 PM, Chad Perrin per...@apotheon.com wrote: Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. A quick search yielded these links: http://wiki.freebsd.org/BuildingFreeBSDWithClang http://www.linuxtoday.com/developer/2009051100335NWBD Interesting, but there is nothing in those that indicates an official direction for FreeBSD. Just that some people are doing some work on it for some reason. Doesn't mean it won't become the official direction at some future time either. Whatever 'official' means. jerry Cheers, Ben ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 08, 2010 at 03:12:25PM -0500, Derek Buttineau wrote: On 2010-02-08, at 2:58 PM, Chad Perrin wrote: Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. Here's last year's status report where they talk about it: http://lists.freebsd.org/pipermail/freebsd-stable/2009-May/049873.html This one looks a little more definitive, but it still uses the term 'exploring' rather than saying it is decided.So, keep listening. jerry It was also in a presentation at BSDCan last year. -- Regards, Derek Buttineau Internet Systems Developer Compu-SOLVE Internet Services Compu-SOLVE Technologies, Inc Phone: 705-725-1212 x255 E-Mail: de...@csolve.net ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 8, 2010 at 2:57 PM, Jerry McAllister jerr...@msu.edu wrote: On Mon, Feb 08, 2010 at 03:12:25PM -0500, Derek Buttineau wrote: On 2010-02-08, at 2:58 PM, Chad Perrin wrote: Can you provide a URL for some discussion of this? I hadn't heard that FreeBSD was moving to Clang. Here's last year's status report where they talk about it: http://lists.freebsd.org/pipermail/freebsd-stable/2009-May/049873.html This one looks a little more definitive, but it still uses the term 'exploring' rather than saying it is decided.So, keep listening. jerry http://www.freebsdnews.net/2009/10/15/clang-llvm-support-on-freebsd/ It's also been talked about on other lists eg current with growing frequency. I seem to remember hearing they hoped for it to be ready as an option for 9.0, although my memory may not be correct on that. -- Adam Vande More ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Monday 08 February 2010 21:19:01 Mihai Donțu wrote: On Monday 08 February 2010 12:54:26 Pieter de Goeje wrote: Even deleting a large file off that raid array I can see a difference, prior to reformatting, i deleted a 190GB file off the raid, under UFS the delete took quite some time (well over 10 seconds), under ext4 the deletion of the same size file took about 3 seconds. File deletion speed is relevant how? It can be, depending on the workload. I (as a Linux user) moved from ext3 to xfs, ignoring the warnings about file deletion [being slow]. Now I _kind of_ regret it. Seems I have more than one program on my laptop that deletes files (kmail's email-expiration thing comes to mind). I also work on a project that creates large log files an deletes them (periodically). When all these programs meet, I go for a coffee. :) I agree that file deletion speed can be important in normal usage scenarios. However my question was asked in the context of an FTP up or download, which does not involve deleting files. :-) -- Pieter de Goeje ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 8, 2010 at 5:46 AM, alex a...@mailinglist.ahhyes.net wrote: I do suspect personally that the ext4 filesystem is the reason for the difference here, since ext4 has a number of features such as deferred disk writes etc. Even deleting a large file off that raid array I can see a difference, prior to reformatting, i deleted a 190GB file off the raid, under UFS the delete took quite some time (well over 10 seconds), under ext4 the deletion of the same size file took about 3 seconds. But what I said with ext4 being faster then the aging UFS still rings true in my mind, look at the recent Phoronix benchmarks for yourself and see (10 pages of benchmarks). http://www.phoronix.com/scan.php?page=articleitem=freebsd8_benchmarksnum=1 (skip to page 7 of the benchmarks if you want to see the I/O stuff relating to disk performance) According to the first page they used the default configuration of all benchmarked OS'es. And what is the default mount option on Linux async The FreeBSD man page for mount describes this async option as follows: async All I/O to the file system should be done asynchronously. This is a dangerous flag to set, since it does not guar- antee that the file system structure on the disk will remain consistent. For this reason, the async flag should be used sparingly, and only when some data recov- ery mechanism is present. The OpenBSD man page has the following additional remark: The most common use of this flag is to speed up restore(8) where it can give a factor of two speed in- crease. Conclusion: you cannot compare filesystem performance, when you give one a unfair speed advantage of what could be a factor two. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
FreeBSD's UFS vs Ext4
Hi Guys, Today I reformatted a machine (network server) thats run FreeBSD nonstop for at least the last 3 years and installed linux on it. I have a raid 0 setup with 2 hard disks in the very same machine. Previously, the maximum I could get across my gigabit enabled network was 60MB/s (megabytes) per second sustained transfer rate. Now that the same machine's raid is formatted with ext4, i am easily sustaining 86MB/s. I cant put it down to the operating system kernel, as to the vast difference in performance, i suspect it is simply ext4 thats producing the better results (I have come to this conclusion because no hardware has changed on that machine, only the OS). So can I safely conclude that ext4 is miles ahead of FreeBSD's UFS performance wise? I'd like to see some feedback.. I am by no means a linux troll. In fact I am far from it. I own many FreeBSD tshirts. I see a number of factors putting freebsd behind: * The teams stubbornness with compiler/base tools (wont move away from gcc 4.2.1 because they just cant accept the GPL2) * The teams stubbornness with the base system binutils (which cause mplayer and other multimedia applications not to build, unless a newer version is installed) * NO interest in developing new filesystems (forget ZFS), i am talking about a base filesystem, ext4 blows the socks off UFS. Using such an old compiler must have a performance impact on the OS. I say this because compilers improve over time, they generate better, tighter, more optimized code. The binutils shipped with freebsd is more than 5 years old now. It's not just my personal test that has shown that linux is ahead in numerous areas (performance wise), but the recent phoronix benchmarks that were released when FreeBSD 8 came out, were pretty damning. I'd like to see what the FreeBSD team has to say on this. Alex ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
Sorry I should clarify that the copy was via FTP to the raid drive in both comparisons; FreeBSD with UFS: Maximum achievable when copying over the network to the raid drive = 60MB/s Linux with ext4: Maximum achievable when copying over the network to the raid drive = 86MB/s Original Message Subject:FreeBSD's UFS vs Ext4 Date: Mon, 08 Feb 2010 01:41:29 +1100 From: alex a...@mailinglist.ahhyes.net To: freebsd-questions@freebsd.org Hi Guys, Previously, the maximum I could get across my gigabit enabled network was 60MB/s (megabytes) per second sustained transfer rate. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Mon, Feb 08, 2010 at 01:41:29AM +1100, alex wrote: Hi Guys, Today I reformatted a machine (network server) thats run FreeBSD nonstop for at least the last 3 years and installed linux on it. I have a raid 0 setup with 2 hard disks in the very same machine. So you had a machine that had run non-stop for 3 years yet you replace the OS. Clever. Previously, the maximum I could get across my gigabit enabled network was 60MB/s (megabytes) per second sustained transfer rate. Now that the same machine's raid is formatted with ext4, i am easily sustaining 86MB/s. I cant put it down to the operating system kernel, as to the vast difference in performance, i suspect it is simply ext4 thats producing the better results (I have come to this conclusion because no hardware has changed on that machine, only the OS). So can I safely conclude that ext4 is miles ahead of FreeBSD's UFS performance wise? No you can't. What about the driver for your NIC? It may be nothing to do with the FS. I'd like to see some feedback.. I am by no means a linux troll. In fact I am far from it. I own many FreeBSD tshirts. Oh well, if you own FreeBSD T-shirts that settles the matter. I see a number of factors putting freebsd behind: * The teams stubbornness with compiler/base tools (wont move away from gcc 4.2.1 because they just cant accept the GPL2) They don't like the license, that's not stubbornness. * The teams stubbornness with the base system binutils (which cause mplayer and other multimedia applications not to build, unless a newer version is installed) Nonsense. * NO interest in developing new filesystems (forget ZFS), i am talking about a base filesystem, ext4 blows the socks off UFS. You say, with your in-depth study of the matter and understanding of filesystems. Using such an old compiler must have a performance impact on the OS. I say this because compilers improve over time, they generate better, tighter, more optimized code. The binutils shipped with freebsd is more than 5 years old now. A codes age has nothing to do with it's performance. It's not just my personal test that has shown that linux is ahead in numerous areas (performance wise), but the recent phoronix benchmarks that were released when FreeBSD 8 came out, were pretty damning. Link please. I'd like to see what the FreeBSD team has to say on this. Alex Despite your FreeBSD T-shirt ownage, your post is a troll. Nobody's interested in your bogus benchmarks opinions on matters that you are not knowledgeable of. Regards, -- Frank Contact info: http://www.shute.org.uk/misc/contact.html ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
Frank Shute wrote: On Mon, Feb 08, 2010 at 01:41:29AM +1100, alex wrote: Hi Guys, Today I reformatted a machine (network server) thats run FreeBSD nonstop for at least the last 3 years and installed linux on it. I have a raid 0 setup with 2 hard disks in the very same machine. So you had a machine that had run non-stop for 3 years yet you replace the OS. Clever. Yes I replaced the OS. Because the box was to also be a PBX (running asterisk, instead of just being a file server/web server for running local web apps). I was continually getting coredumps with asterisk. After filing numerous bug reports and hitting dead ends with the asterisk devs, I had enough, because none of them knew how to debug the problem under freebsd, I got fed up and moved the box over to linux, and to my surprise, no more core dumps. I see a number of factors putting freebsd behind: * The teams stubbornness with compiler/base tools (wont move away from gcc 4.2.1 because they just cant accept the GPL2) They don't like the license, that's not stubbornness. Wow thats a good reason to use ancient compilers and assemblers. * The teams stubbornness with the base system binutils (which cause mplayer and other multimedia applications not to build, unless a newer version is installed) Nonsense. You dont see having a set of binutils thats not SSE3 or SSE4 capable as a problem? It's nonsense? Using such an old compiler must have a performance impact on the OS. I say this because compilers improve over time, they generate better, tighter, more optimized code. The binutils shipped with freebsd is more than 5 years old now. A codes age has nothing to do with it's performance. Clearly you know nothing about how compilers generate and optimize code. If this isnt a problem, why would new versions of gcc and binutils continue to surface. Well I can see three obvious reasons, improved code generation, bug fixes, new features. It's not just my personal test that has shown that linux is ahead in numerous areas (performance wise), but the recent phoronix benchmarks that were released when FreeBSD 8 came out, were pretty damning. Link please. Sure, no problem, enjoy: http://www.phoronix.com/scan.php?page=articleitem=freebsd8_benchmarksnum=1 Go on, I am waiting for you to poke holes and attempt to totally invalidate those benchmarks too. I'd like to see what the FreeBSD team has to say on this. Alex Despite your FreeBSD T-shirt ownage, your post is a troll. Nobody's interested in your bogus benchmarks opinions on matters that you are not knowledgeable of. Regards, I guess you cant see the difference between a troll and a complaint. I have been using freebsd since the 4.x days. It seems you have quite a chip on your shoulder, frank. Alex. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
On Sunday 07 February 2010 15:41:29 alex wrote: Hi Guys, Today I reformatted a machine (network server) thats run FreeBSD nonstop for at least the last 3 years and installed linux on it. I have a raid 0 setup with 2 hard disks in the very same machine. Previously, the maximum I could get across my gigabit enabled network was 60MB/s (megabytes) per second sustained transfer rate. Now that the same machine's raid is formatted with ext4, i am easily sustaining 86MB/s. Too many variables. The difference in performance could be due to: 1) Slow filesystem. 2) Slow network (nic). 3) Slow FTP server. The fact that the limit is 86MB/sec (which is very low for a raid0 array) makes me think the box suffers from sub optimal network performance during a simple stream test like yours. This could be due to FreeBSD having a poor network driver for your particular NIC or could be due to insufficient tuning of the TCP parameters for this particular test. You haven't given any details about the hardware, network tuning done, how you configured the filesystem, raw filesystem performance, raw network performance. If you want a meaningful response based on more than guesswork, please gather more data. -- Pieter de Goeje ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
Hi, On 07 February 2010 pm 22:41:29 alex wrote: Today I reformatted a machine (network server) thats run FreeBSD nonstop for at least the last 3 years and installed linux on it. I have a raid 0 setup with 2 hard disks in the very same machine. can you do the same for FreeBSD? Just install 8.0 and run the same test as before. As the machine running for three years, it could be that the wrong driver for some device was used. I would never compare a running system directly to a freshly installed one. Erich ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: FreeBSD's UFS vs Ext4
Pieter de Goeje wrote: The fact that the limit is 86MB/sec (which is very low for a raid0 array) makes me think the box suffers from sub optimal network performance during a simple stream test like yours. This could be due to FreeBSD having a poor network driver for your particular NIC or could be due to insufficient tuning of the TCP parameters for this particular test. Hi Pieter. You are right about there being a number of possibilities, however: *The same machine, which over the years has had a number of revisions of freebsd on it (have buildworlded the thing from 7- 7.1 - 7.2 - 8), the performance was always roughly the same amongst the versions, I dont agree with the possibility of the ftp server being 'slow' as I am the only person who copies data to that machine, and the machine is always under a very low (almost non existent) load. * Network card is an Intel Pro 1000, on the server. This is a PCI card (not pci-e), so I believe PCI bus bandwidth limitations may be responsible for me not being able to achieve the maximum 100MB/s network rate (as you mention that 86MB/s is slow for raid0) * The intel network card driver on freebsd and linux are both fairly rock solid and well written. I dont see it being an issue with NIC drivers (they are not vastly different). * Both OS's were stock standard installs, no jumbo frames enabled, no fiddling with sysctl network values. I am happy with 86MB/s anyway, It's a huge improvement of the 60MB/s barrier I could never get past when that machine was running FreeBSD. To get the rest of the speed, I'd probably have to install a pci-e card on the server. I do suspect personally that the ext4 filesystem is the reason for the difference here, since ext4 has a number of features such as deferred disk writes etc. Even deleting a large file off that raid array I can see a difference, prior to reformatting, i deleted a 190GB file off the raid, under UFS the delete took quite some time (well over 10 seconds), under ext4 the deletion of the same size file took about 3 seconds. But what I said with ext4 being faster then the aging UFS still rings true in my mind, look at the recent Phoronix benchmarks for yourself and see (10 pages of benchmarks). http://www.phoronix.com/scan.php?page=articleitem=freebsd8_benchmarksnum=1 (skip to page 7 of the benchmarks if you want to see the I/O stuff relating to disk performance) ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org