“The journaling operation itself does impose a performance penalty on disk writes. Mac OS X Server alters the sizes of certain buffers used for file transactions when journaling is enabled, which mitigates much of the performance hit, reducing it from the 10-15 percent range down to the 2-5 percent range, for a system with 512MB of RAM. The more RAM you have, the more buffering can be used, so your performance hit decreases accordingly. This buffering does not occur on “regular” OS X, which is one reason why Apple is not supporting or recommending its use on non-OS X Server systems.” Read the article at WorkingMac.
MacOSX Journaling: What it is, who Needs it
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
There’s an interesting thread on performance of journaled FSes at http://arstechnica.infopop.net/OpenTopic/page?a=tpc&s=50009562&f=48…
Thanks for pointing this one out – I suspected as much (the buffering offsetting performance hits) for OS X Server, but I couldn’t confirm it.
And ignore the unenlightened folks who comment on OSNews’ bias. The bias is there – this site is slanted toward providing operating system/programming news
Peace,
‘Rithm
Yikes. That performance is pretty dismal. My system: 2GHz P4, Linux 2.5.46 640MB DDR-266 RAM, 4800 RPM IBM 20GNX *laptop* hard drive (XFS):
274 seconds total
149 seconds of transactions (134 per second)
53.53 megabytes read (200.04 kilobytes per second)
170.80 megabytes written (638.31 kilobytes per second)
G4-450 OS X 10.2.2 512MB RAM, 80GB 7200 RPM Western Digital hard drive (HFS+ journaling):
481 seconds total
419 seconds of transactions (47 per second)
53.53 megabytes read (113.95 kilobytes per second)
170.80 megabytes written (363.61 kilobytes per second)
The funny thing is that my hardware is hopelessly outclassed. Laptop hard drives can generate decent transfer rates (20MB/sec, comparable to an older 7200 RPM 20GB desktop hard drive) but because of their slow rotational speed, have terrible transaction performance. Still, all that devel-kernel + XFS goodness managed to spank not only OS X using a much faster, much newer desktop hard drive, but it managed to beat a couple of 10K RPM SCSI systems as well
And pay attention to the exact commands people used when setting up this benchmark. The standard policy seemed to be to set “number of files” and “number of transactions” to 20,000. Some people messed with block sizes and stuff and ended up getting some rather shady numbers.
We wouldn’t need a journaling FS if OSX didn’t crash so much. I though its unix design was supposed to make it stable? I keep getting that message that says i must restart my computer, and i have a brand new system. Journaling would speed up the reboots, so i’m interesed, but will i notice the performace penalty? If so, how bad is it?
>We wouldn’t need a journaling FS if OSX didn’t crash so much.
I am the one who always cries as to how slow OSX is. But I have to say one thing: MacOSX has crashed on me only ONCE and never gave me any real stability problems (note: I am not running OS9 apps).
I suggest you completely re-install OSX, as obviously something is wrong on your setup. Or, it could be an application you installed that creates the problem and not the OS at all.
You are telling us a 2 ghz P4 runs twice faster than a g4 450 Mhz with less/slower ram ? I would not call this abysmal performance (unless you are speaking about the PC laptop).
As for the need of journaling, I have been running a 240 gig raid 0 on OSX since 10.0, I did experience a few crashes since then (I can count them on my hands) but the fs integrity/checkup at startup never took more than a few minutes. Maybe the journaling makes sense for critical data and terabyte systems, but even for powerusers it is overkill. It will not save you from a hardware drive failure and that is where the danger resides.
I have to also take MacOSX’s side to this. When my G4 Cube had some hardware problems, it would randomly reboot. I had comment to it back then, and I did say that I was afraid to turn on the Cube, just in case the fs will go berzerk. But OSX itself never failed me on the integrity part. When OSX would come up again from the random reboot, ALL the data was intact with no problems at all.
Since I fixed this (hardware) problem I have only seen a single “real” software crash from OSX, since June 2002. MacOSX is dog slow, but it is not unstable in my experience (again the note: I do not run OS9 software).
Its funny that you complain of crashes. OS X rarely crashes on my system and I’m running a 400 G3. Are you using MS products, they are about the only thing thats crashed my system…and a shareware I downloaded.
Most common cause: Bad RAM. Wish OS X had a good utility for testing that ….
see recent RAM thread on http://www.macintouch.com
It takes very little CPU to drive a hard drive to its knees. A hard drive can transfer at most 30 MB/sec and make an I/O once every 10ms or so. An old Pentium 100 can handle that! This benchmark is entirely I/O bound, and as a result (try reading the rationale for setting the parameters as they did!), the only thing that matters is how fast the disk is. And my 4800 RPM laptop hard drive with XFS trashed a 7200 RPM desktop hard drive with HFS+. I’d call that dismal performance.
Hey,
You might be right, there. However, keep in mind that there are different disk controllers in use, and the speed with which the disk controllers are fed information is different between the machines. Even the delivery of instructions TO the controller is along a different path. There’s a lot more going on than just the hard drive performance. Which opens another can of worms, actually…
Peace,
‘Rithm
*shakes head incredulously that buying a laptop from a grocery store was a GOOD idea*
*hugs iBook*
*warms hands over Medion’s heat exhaust vents*
“We wouldn’t need a journaling FS if OSX didn’t crash so much. I though its unix design was supposed to make it stable? I keep getting that message that says i must restart my computer, and i have a brand new system. Journaling would speed up the reboots, so i’m interesed, but will i notice the performace penalty? If so, how bad is it?”
You could have bad ram. There is some good information at macintouch.com (http://www.macintouch.com/badram01.html )regarding bad ram. Seems like in most cases, removing and/or changing the bad ram fixed the problem.
– Mark
Yes, Eugenia is right, you should not be having experiences like that with OS X. However, Mark has hit a good point. You can have zero problems with OS 9, but big problems with OS X if you have added “cheaper” RAM that OS X doesn’t like. The start-up post won’t detect any problems, so it’s really a killer if you don’t know. They suggest getting really good add on RAM like Kingston, but I’ve always had good luck with getting RAM for Macs through Crucial.
But, it might be good to do what Eugenia suggested first, to help make sure that there may be a hardware problem, rather than a software problem.
Journaling is not just for power users or critical systems.
If you are in the middle of overwriting a file and your system crashes, then you will lose both versions of the file. Unless you have a journaling filesystem, in this case you will only lose the new version of the file. The old one will remain.
This fact alone is enough that everyone should be running a journaling filesystem, IMO.
I’ll look into the RAM, but for two months i was useing a stick of 512MB PC2100 DDR that it came with, and it crashed the first or second day i had it. It used to freeze up for long periods of time, but that hasn’t happened in a while. Also, the sleep mode kicks in too frequently and takes too long to wake up. I click something, i get the ball, have to wait about 6 seconds, hard drive starts making some noise. Then it works. Is there anyway to make it not do that?
Jay,
What type of machine is it? What OS Rev did it come with, and what OS version is it running at now? You mentioned it had a single stick of RAM when you got it – had you added more peripherals, or were/are you running it stock? It shouldn’t be performing poorly, so let us know and the folks here can help you out. Also, get the RAM testing utility – the RAM may have been/may be compromised. I don’t see that being the problem, but it’s nice to cover all the bases, right?
If this is too far OT, please moderate me down – Jay, you can send me any info that can’t get posted here to the above ‘mail and hopefully take care of your problems
Peace,
‘Rithm
This is what causes the mouse cursor and the text cursor to go so slowly. With only 512MB RAM, journaling is a big hit on the OS.
So I have to upgrade to 1GB (or more) RAM. Fortunately that’s cheap compared to the cost of the machine.
Sometimes I wonder how Apple managed to hire only engineers who know how to write slow code. It used to be that Apple wrote efficient code… Apple ][, original Mac. Now, it is all bloatware that needs gigs of RAM, fast hard drives, multiple processors, etc, just to work at a speed that is usable. I wonder if it is because more and more of the OS is being written in Java…
(shiny white)