Ubuntu 10.04 filesystems and boot times

I haven’t really mentioned it much, but the few times in the past month that I’ve used Ubuntu 10.04 have been rather disappointing. I haven’t dared install it on anything except this machine, and in most cases that’s been only borderline acceptable.

Disk access seems to be one of the biggest problems for me. The time it takes to start up, load programs, update software and make file transfers is exceptionally bad, on a machine I consider to be extremely fast. In fact, most times I get a more responsive desktop when using the live system over USB2.0, as opposed to a native install.

In situations like that, I usually blame the filesystem. Ubuntu uses ext4 by default, which has the reputation of being rather fast; I use ext2 by default, which I believe to be a better solution for low-, low-end hardware. On a fast machine it should be even speedier.

(Yes, I know, ext2 means no journalling, which means I might lose data if something goes wrong. I know. Everybody has their favorite filesystem, and everybody lost data to the other ones. Spare us the war stories. πŸ‘Ώ )

Here’s a look at boot times, which can be measured precisely, and suggest how well the system is handling disk access … or at least, the effect the filesystem might be having on system performance.







In many ways, there’s not much here that will shock and awe. ext2 is the fastest by a hair, with ext3 and 4 in close quarters, and reiserfs doing well too. jfs is far back there in last place, and xfs … well, xfs is xfs.

I also gave each installation the benefit of the doubt, allowing it to boot twice or even three times before claiming a bootchart, in case there were first-run tasks to be written. Each system was fully updated, but there were no tweaks added — not even the classic noatime adjustment.

Take away from this what you will; for me, it told me that the filesystem wasn’t the source of my problems, and that I didn’t need to micromanage the installation to find the source of my disk woes. I will be sticking with ext2 in spite of all the fear and loathing that early filesystems attract. But I’ll probably be sticking with Arch Linux too. 😈

P.S.: Yes, yes, btfrs, blah, blah, blah. πŸ™„

20 thoughts on “Ubuntu 10.04 filesystems and boot times

  1. Anthony

    i thought journaling was the whole point of ext4 ?
    and don’t you mean “stickin with ext4 , rather than 2, since that’s not very knew is it πŸ™‚ or since the link, mebe 3 lol

    2 is old, 3 aint that new, 4 is the new kid on the block.

    how about pushing the boat out and getting ZFS on that .

  2. Dennis Decker Jensen

    Why do you think ext2 is less safe than ext3 or ext4? As far as I know, the primary reason, why one is using ext3 or ext4, is that it saves boot time of the fsck that ext2 needs occasionally. Ext2 isn’t any less safe than ext3, which is exactly ext2 with added journaling. Ext4 may have added some features improving the safety over ext2. I don’t know.

    1. Nugnuts

      Ext2 isn’t any less safe than ext3, which is exactly ext2 with added journaling.

      The added journaling is exactly what makes ext3 safer. The journaling makes corruption less likely.

  3. scott

    my ubuntu 10.04 test was rather slow as well on this old laptop. Slow at everything, its slower then its previous versions. I recently had arch linux with openbox on it so that may be why it seems slower to me haha.

  4. Nugnuts

    A more responsive desktop from a live system rather than a native install? Sounds like something is wrong, even if the boot charts don’t reveal any specific issues. Have you seen similar problematic behavior with other distros on the same machine since you’ve seen this happening with Ubuntu 10.04? Any indication of disk problems? Is DMA not being turned on for some reason or some such?

  5. peterix

    No matter the filesystem, all the current kernels are absolutely and completely broken (whole range between 2.6.18 – 2.6.33). They have bad, bad problems with USB transfers and general IO lockups (I can’t use the term IOwait for something that makes my machine unresponsive for half a minute with a damn Intel SSD installed). Overuse of disk cache while migrating application binaries to the swap seems to be the most common problem. Hell, I have this problem with 8GB of RAM installed.

    Yes. the kernel is broken. It’s been that way for while now. By some sheer luck or rather /misfortune/, not enough people are screaming bloody murder yet. AFAIK, if you run RAID or LVM, it doesn’t seem to be that bad… maybe that’s the reason.

    1. Nugnuts

      FWIW, I’m not using RAID, LVM, nor an SSD drive and have had no performance problems with any of the latest kernels.

      1. peterix

        You must be pretty lucky then πŸ™‚ Ive got three machines here that have these problems.

  6. John Bohlke

    I know that I have seen a lot of issues with io choking when I am using Gnome and KDE4. If I switch to other windows managers, xfce, openbox, etc, the problem is mitigated. So I think it isn’t the kernel, it is problem something in the virtual file systems that Gnome and KDE4 are using. Just a possibility.

  7. Luca

    Erm… I may just be being dumb, but ext4 took 33.31 secs to boot and ext2 took 33.48. That means ext4 is faster, contrary to what you said in your post. Right? πŸ˜›

    It is entirely possible that USB will be faster than a laptop HDD, mainly due to random access being faster. The actual transfer speeds between the two are usually about the same as well.

    Have a look at these two articles from 2008. The flash drive tops out reads of 28Mb/s, and the average of the slowest (of the high-end range at that time – thats a mouthful) is 37Mb/s. However the real key at bootup is random access, hence the flash drive wins.


    One thing you might want to do is to upgrade to a newer, faster, HDD. I upgraded to a 7200rpm drive, and the speed increase is definitely noticeable, fortunately the affect on battery life and noise isn’t πŸ™‚

  8. Pogeymanz

    These issues don’t test filesystem speed or responsiveness or IO speed. Bootcharts are confounded by the mount-time required for certain filesystems. If I remember correctly, it is known that JFS takes a long time to mount, but that doesn’t mean it’s necessarily a slow filesystem.

  9. Matt

    Ubuntu gets slower with each release.
    It’s pretty much unusable for me now on my old AMD3200 system.
    What the heck have they done to this distro?

    1. mulenmar

      AMD 3200 — isn’t that a 2GHz 64-bit CPU? I’m surprised that even Ubuntu could be slow on that…

      Then again, it’s been a long time since I used Ubuntu on a single-core system. *eyes the HP Pavilion a1000 he’s Frankenstein-monster’d to life with a Mini-ATX PSU* I’ll get back to you on that, though.

      1. mulenmar

        Just tested Xubuntu 10.04 i386, and there doesn’t seem to be any slowness. I just created two partions, / and swap, and used ext4 for the /. No other speed tweaks.

        It still used just over 250MB of memory after a fresh boot-up, though. Standerd Ubuntu uses well over 350MB on my netbook, too, so could you check how much RAM your machine has? Lack of memory may be causing the slowness.

  10. Mark

    I don’t see what the fuss is about.
    My Ubintu Server takes Five Minutes to boot from a USB key.

    I have 4 x 2Tb disks in a mdraid 5, so used a /boot on usb and a root and data raid5 on disk.

  11. James

    EXT4 can be created without a journal.

    mkfs.ext4 -O ^has_journal /dev/sdX

    And EXT2 can be mounted as EXT4.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s