On Thu, Aug 28/97, Nigel Metheringham <Nigel.Metheringham@???> wrote:
>
> I must admit I find it had to understand why people use lsearch
> extensively - for a partial search you can be doing a lot of file scans.
> Personally I have *everything* as dbm files.
Well, I do it mainly because I don't think my files are big enough to
gain a lot of performance advantages yet. Even though I have a pretty
extensive list, it's only 14k and 770 entries. It's almost guaranteed
it's already in buffer cache most of the time as there's lots of exim's
running most of the day. Is the time to scan under 1000 entries really
significant enough that it would outweigh what ever overhead there is in
using dbm? At least to start with I would be using ndbm as our platform
is Solaris.
I suppose I should try in one our main mail server and see if the load
average drops significantly. Since I managed to block almost all the
relay spammers who were using our machine I must say that the load
average on the machine is relatively low, running around 3-5 most of the
day and dominated by the pop mail server and local delivery program from
what I can see.
The machine is doing a little over 100K messages input daily with 98% in
the queue under a minute locally and 93.5% for the messages with at
least one remote delivery.
--
Artificial Intelligence stands no chance against Natural Stupidity.
GAT d- -p+(--) c++++ l++ u++ t- m--- W--- !v
b+++ e* s-/+ n-(?) h++ f+g+ w+++ y*
