Re: [Exim] Hashing level 2.

トップ ページ
このメッセージを削除
このメッセージに返信
著者: Yann Golanski
日付:  
To: Guillermo Llenas
CC: exim-users
題目: Re: [Exim] Hashing level 2.
--
Quoth Guillermo Llenas on Mon, Sep 08, 2003 at 11:00:27 -0300
>     for example, if the spool directory is /export/mail, the spool
> file for user pepe@??? would be
> /export/mail/mydomain.com/p/e/pepe , this is using directories equal to
> domain and the first letter and the second letter (if any). This is
> hash-spool=2.


This is a terrible way of doing things. Names are not evenly spread by
the alphabet and you will end up with directories empty and some over
full.

There is some code in Exim about a hashing algorithm which will return
you a mod/div set -- nhash to be exact. So, your directories will look
like: /var/spool/mail/[0-9].*/mydomain.com/[0-9].*/[0-9].*/username.
This way, with good choices of numbers, you have a well spread hashing.
Now, you can take those an mount then on different disks/NFS/etc and be
sure that no system is going to get swamped.

If you check the white paper I wrote on Exim a long time ago, you'll see
a good enough hashing: http://gridlock.york.ac.uk/~yann/lsm.pdf

If you insist on using the above, then look at the string function.
Again, I would not recomand using this as it will lead you into trouble.
</cassandra>

>     Any help will be thanked for eternally :P


Hope this helps.

--
yann@???                  -=*=-                      www.kierun.org
    PGP:   009D 7287 C4A7 FD4F 1680  06E4 F751 7006 9DE2 6318
--
[ Content of type application/pgp-signature deleted ]
--