[dokuwiki] Re: The database question: a new spin

> Andreas Trawoeger wrote:
>
> Just out of curiosity: Who is running a Dokuwiki Site where performance does 
> matter?
> 
> Most important thing I'm using a dokuwiki isn't performance it is 
> availability.

THANK YOU!  I've been talking about this thing I called "peak load performance" 
and "server capacity" for the past week or so when the real word I wanted was 
AVAILABILITY.  Thank you!

Improving availability is the main reason for allowing database underpinnings.  
You can improve individual page load performance too, but only if the database 
"cache" has been built.

A database server basically gives you two benefits:

(1) You get a battle-hardened, smartly optimized indexing system.
(2) You get in-memory caching across multiple page requests.

Without a database you're using more clock cycles and likely more memory 
finding information, and you're hitting the slow, head-limited disk to get 
everything you need, possibly multiple times.  Both translate to more resources 
consumed per page request out of a server's finite, predetermined resources, 
and thus a lower peak load and a lower threshhold for unavailability.

> So if there is any need for improved performance I would first write a 
> performance
>benchmark, test how dokuwiki behaves with different filesystems, start giving
>performance tips and start looking for performance bottlenecks in dokuwiki.

And I would too.  We're on the same page.  I suggested this in an earlier email.

> Before I start heading the database way, because performance 'could' be 
> faster.

You're right.  Single page load times 'could' be faster, but availability MUST 
improve.  If you reduce resource consumption, the server's capacity must go up. 
 Availability is the word I meant.

> The only area where I would like a database is user management. But I think it
>would be easier to improve dokuwikis authentification via LDAP or integration 
>in
>other CMS then writing yet another user management system in MySql.

I agree that a database would do a better job of managing large sets of 
orthogonal but related information, but I hate the idea of having to back up 
both a file system and a database to back up my server.

On the other hand, we're on the same page when you say that we could allow for 
integration with other systems, provided that DokuWiki doesn't require that 
integration out of the box.  In practice this just means being forward-thinking 
with DokuWiki design as we go.

~joe

--
DokuWiki mailing list - more info at
http://wiki.splitbrain.org/wiki:mailinglist

Other related posts: