I would really like to see expire gone

[Alex Kiernan]

davidsen at tmr.com (bill davidsen) writes:

> In article <72ptqd8wkc.fsf at demon.net>, Alex Kiernan  <alexk at demon.net> wrote:
> 
> | OK, I found a bigger box to try on (dual 1GHz UltraSparc III w/ 8GB of
> | RAM)
> 
> That's a bigger box indeed, and perhaps a bigger box than many people
> have for news. However, if this code actually turns into something which
> might be reasonable to try on more modest machines, I'd love to see how
> it goes on an x86 box with 2 or 4GB RAM.
> 

If anyone wants to try, you're welcome to the code - you just need to
supply a history file.

> | I can insert all 46M articles in the history in 20 minutes (with a
> | 3.95GB cache), unfortunately thats about as far as you could take it
> | entirely in core as Berkeley DB 4.1.25 is limited to a max of 4GB of
> | cache AFAICT (even if you build it 64 bit).
> 
> That still sounds as if it would be practical in terms of keeping up
> with current loads.
> 

I'd agree, but I think switching INN to something where you can
parallelise article reception with getting stuff to disk is
essential... when it has to go to disk it can still keep up with the
article rate (even when running with logging & transactions on), but I
doubt you'd get the articles received and stored to disk at a
sufficient rate.

-- 
Alex Kiernan, Principal Engineer, Development, THUS plc