blog.6admin.io
guest@6admin.io:~$
filesystem, linux

ext3: accelerate your disk accesses

The ext3 filesystem is the default on every distribution for years now and it is mounted with the slowest possible options. Three words to add to the fstab and your disk accesses are halved.

The three options

/dev/sda1  /  ext3  defaults,noatime,data=writeback,commit=60  0  1

noatime. The access time (atime) is updated at EVERY file read. A cat file updates the atime of file AND of every directory in the path. On a web server reading thousands of files per second, atime is an invisible disk overwrite. noatime removes it outright. Cost: ls -lu no longer shows the last read time, which has never been useful to anyone.

data=writeback. The default journal mode is ordered: data is written after the journal. writeback reverses the order: data is written directly, the journal only tracks metadata. The result is faster, but in case of crash, the file content may be stale while the file size is correct. For a critical data server, stay on ordered. For a web server serving static files, writeback is risk-free.

commit=60. The journal is flushed every 5 seconds by default. commit=60 spaces it to 60 seconds. Less journal IO, more data in cache, latency halved. In exchange, a crash within the 60-second window can lose that window’s data.

The hierarchy

If you really want to accelerate, the hierarchy is:

  1. RAM: tmpfs for temporary files (the fastest disk cache is the RAM disk)
  2. RAID 0 or 10: parallel writes across multiple spindles
  3. The FS: the options above
  4. The disk: a 15k RPM beats a 7.2k RPM, always

The order of operations: start with noatime (free, risk-free), then data=writeback (test in staging), then commit=60 (owned). Never touch RAID geometry to compensate for a badly configured FS.

The test

Before/after, the simplest:

# Before
time find /var/www -type f | wc -l

# After remounting with noatime
time find /var/www -type f | wc -l

On a directory of 50,000 files, the difference is typically 2x. On a mail server with tens of thousands of files per Maildir, it is the same ratio.

[edit: 2026] ext3 has been replaced by ext4 (which integrates the optimizations by default) then by XFS and Btrfs. But noatime remains the highest-return option on any Linux filesystem, ext4 included: atime is still the invisible overwrite that costs a disk IO on every read. The rest has evolved, noatime is eternal.