4 ms·

You can't specifically access the cached version, but the OS will (assuming the file isn't too large) remember the blocks read from disk in the cache for a whil

You can't specifically access the cached version, but the OS will (assuming the file isn't too large) remember the blocks read from disk in the cache for a while (until the memory is needed by something else) so that next time the file is ready those blocks do not need to be read from disk. This can make a huge difference on traditional drives. /dev/null is the abyss, but the cat command will read all the blocks comprising the file from the filesystem before sending them to that abyss. You can see the effect like so: dspillett@swann2:/tmp$ time cat a.bigish.file > /dev/null real 0m3.830s user 0m0.000s sys 0m0.468s dspillett@swann2:/tmp$ time cat a.bigish.file > /dev/null real 0m0.165s user 0m0.004s sys 0m0.164s dspillett@swann2:/tmp$ time cat a.bigish.file > /dev/null real 0m0.148s user 0m0.000s sys 0m0.144s (if "a.bigish.file" has only just been created, it will likely already be in RAM, run sync; echo 3 > /proc/sys/vm/drop_caches as root or other privileged user to clear the not-currently-active cache & buffers, or chose an older file) (note though that this clears all cache machine-wide, not just cached information from your current session)

Foivos 6y ago
Thank you for the detailed explanation. So any interaction with the file would put it in memory will do. Of course assuming that you interact with the whole file. So if you have a big file and use "head", I guess it will not be enough. But you use "less" and scroll the whole thing you achieve the same result as in the command above right?
- dspillett 6y ago
  Basically, yes. There are some complications: * If the file is large then blocks at the start will be dropped from cache as you move further in, or will quickly be pushed out by IO activity elsewhere on your system. * Some commands open files with options that inhibit caching, so not everything will have this effect. > Of course assuming that you interact with the whole file. Reading some of the file will have the effect for that portion of the file, so if subsequent commands only touch the same parts then the effect is the same.
  - Foivos 6y ago
    Great explanation! Thank you!

Foivos 6y ago

Thank you for the detailed explanation. So any interaction with the file would put it in memory will do. Of course assuming that you interact with the whole file. So if you have a big file and use "head", I guess it will not be enough. But you use "less" and scroll the whole thing you achieve the same result as in the command above right?

dspillett 6y ago
Basically, yes. There are some complications: * If the file is large then blocks at the start will be dropped from cache as you move further in, or will quickly be pushed out by IO activity elsewhere on your system. * Some commands open files with options that inhibit caching, so not everything will have this effect. > Of course assuming that you interact with the whole file. Reading some of the file will have the effect for that portion of the file, so if subsequent commands only touch the same parts then the effect is the same.
- Foivos 6y ago
  Great explanation! Thank you!