lkml.org 
[lkml]   [2001]   [Mar]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Hashing and directories
Pavel Machek wrote:

> Hi!

> > I was hoping to point out that in real life, most systems that
> > need to access large numbers of files are already designed to do
> > some kind of hashing, or at least to divide-and-conquer by using
> > multi-level directory structures.

> Yes -- because their workaround kernel slowness.

Not just kernel ... because we use NFS a lot, directory searching is
a fair bit quicker with smaller directories (especially when looking
manually at things).

> I had to do this kind of hashing because kernel disliked 70000 html
> files (copy of train time tables).

> BTW try rm * with 70000 files in directory -- command line will overflow.

Sort of my point, again. There are limits to what is sane.

Another example I have cited -- our ticketing system -- is a good one.
If there is subdivision, it can be easier to search subsets of the data.
Can you imagine a source tree with 10k files, all in one directory? I
think *people* need subdivision more than the machines do, a lot of the
time. Another example would be mailboxes ... I have started to build a
hierarchy of mail folders because I have more than a screenful.

> Yes? Easier to type cat timetab1/2345 that can timetab12345? With bigger
> command line size, putting i into *one& directory is definitely easier.

IMO (strictly my own) it is often easier to have things subdivided.
I have had to split up my archive of linux tarballs and patches because
it was getting too big to vgrep.

> > A couple of practical examples from work here at Netcom UK (now
> > Ebone :), would be say DNS zone files or user authentication data.
> > We use Solaris and NFS a lot, too, so large directories are a bad
> > thing in general for us, so we tend to subdivide things using a
> > very simple scheme: taking the first letter and then sometimes
> > the second letter or a pair of letters from the filename. This
> > actually works extremely well in practice, and as mentioned above
> > provides some positive side-effects.

> Positive? Try listing all names that contain "linux" with such case. I'll
> do ls *linux*. You'll need ls */*linux* ?l/inux* li/nux*. Seems ugly to
> me.

It's not that bad, as we tend to be fairly consistent in a scheme. I
only have to remember one of those combinations at a time :)

Anyway, again I apologise for starting or continuing (I forget which)
this thread. I really do understand (and agree with) the arguments for
better directory performance. I have moved to ReiserFS, mainly for the
avoidance of long fsck (power failure, children pushing buttons, alpha
and beta testing of 3D graphics drivers). I *love* being able to type
"rm -rf linux-x.y.z-acNN" and have the command prompt reappear in less
than a second. I intended merely to highlight the danger inherent in
saying to people "oh look you can put a million entries in a directory
now" :)

*whack* bad thread *die* *die*

> Pavel

--
/* Bill Crawford, Unix Systems Developer, Ebone (formerly GTS Netcom) */
#include <stddiscl>
const char *addresses[] = {
"bill@syseng.netcom.net.uk", "Bill.Crawford@ebone.com", // work
"billc@netcomuk.co.uk", "bill@eb0ne.net" // home
};
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:22    [W:0.127 / U:0.296 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site