Announcement

Collapse
No announcement yet.

IBM Baking Some Nice Optimizations To EXT4's Multi-Block Allocator

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by sebastianlacuesta View Post

    Back in the day, having a 233Mhz Pentium 2, ext3 ate most of the CPU cycles while reiserfs worked like a charm.
    Yeah, ext3fs had horrible performance to the point that quite a few users, including companies like Google, refused to use it. Google famously stuck with ext2 till ext4 became stable enough to use. I know a lot of mere mortals simply switched to ReiserFS while others used JFS or those that were very sure of their underlying hardware (by which I mean they used SCSI and had reliable drives and UPSes) used XFS or stuck with ext2. I used ReiserFS when I had a lot of small files and other FSes when Reiser wasn't optimal but avoided ext3. I remember using JFS for /home and / and ext2 for /boot with ReiserFS for /tmp, code hosting filesystems, and mail spools. XFS was pretty bad performance wise with small file deletes in large directories back then. It was optimized for large contiguous files like video and it had cases where power outages could result in data corruption (hence the UPS being highly recommended on XFS systems).

    Comment


    • #12
      Originally posted by luno View Post
      this look great, which filesystem perform well with small files ?
      ZFS can store small files (e.g. ~220 bytes) in block pointers, which avoids allocating data blocks for them.

      Comment


      • #13
        stormcrow Why would you use Reiserfs for /tmp?
        Does Linux not have tmpfs back then?

        Comment


        • #14
          Originally posted by ryao View Post

          ZFS can store small files (e.g. ~220 bytes) in block pointers, which avoids allocating data blocks for them.
          Btrfs has similar techniques. Chris Mason was influenced by ReiserFS, which he maintained.

          Comment


          • #15
            Originally posted by luno View Post
            this look great, which filesystem perform well with small files ?
            btrfs

            Comment


            • #16
              Originally posted by luno View Post
              this look great, which filesystem perform well with small files ?
              Tmpfs, it has the best performance LOL

              Just joking, while tmpfs is indeed a filesystem it cannot persist your data.

              Comment


              • #17
                Originally posted by NobodyXu View Post
                stormcrow Why would you use Reiserfs for /tmp?
                Does Linux not have tmpfs back then?
                yes, it had, but sometimes you need a large /tmp and RAM was a scarce resource (I'm not the original poster, but maybe he had the same constraints).




                Comment


                • #18
                  Originally posted by cynic View Post
                  yes, it had, but sometimes you need a large /tmp and RAM was a scarce resource (I'm not the original poster, but maybe he had the same constraints).
                  Yes, that's true, but with zswap or zram I think this is less of a problem since they can be compressed.
                  They can also be swapped out if needed.

                  Comment


                  • #19
                    Originally posted by NobodyXu View Post

                    Yes, that's true, but with zswap or zram I think this is less of a problem since they can be compressed.
                    They can also be swapped out if needed.
                    this is new stuff... we're talking about early 2000s! 😅

                    Comment


                    • #20
                      Originally posted by cynic View Post
                      this is new stuff... we're talking about early 2000s! 😅
                      Oh right, I lost the context.

                      Comment

                      Working...
                      X