Re: Parallel delivery rates?

[Michael Loftis <mloftis@wgops.com>]

Matti Aarnio wrote:
> 
> On Thu, Aug 02, 2001 at 10:49:17PM -0700, Michael Loftis wrote:
> > OK I've noticed that zmailer tends to only get up one delivery thread
> > per domain...  What exactly controls this behavious?  I've tried
> > danged-near everything under the sun to up those limits (I have managed
> > to apparently get slightly more deliveries going in parallel on the
> > average though which does help)...
> 
>    It somewhat depends on the version -- very early versions (2.2.*)
>   do schedule parallel threads, their scheduling model is different.
>   The "parallel threads" support has existed in the new one for about
>   three years (since 1998, I think), although even I haven't used it
>   all that much having no usual need for it.
> 
>    Have you read the initial (lengthty) comments in the beginning
>   of the boilerplate   scheduler.conf(.in)   file ?
> 
>    There I try to illuminate the relationship of the different control
>   variables.
> 
> > Really it seems that only one process (ta?) is delivering to a
> > host/domain.  The manual is never really clean and clear on the whole
> > thread, process, job, channel relations....
> 
>    Which of the manuals ?  The big bulky SGML/HTML thing ?
>   The scheduler.8,  or something else ?  (Of which version of ZMailer ?)
>   It is entirely possible that I haven't gotten around to rewrite the
>   texts about scheduling in the Manual.
> 
>    In the   scheduler.8  there is this detail:
> 
>        maxthr (1)
>               This limits the number of parallel transport agents
>               within  each  thread;  that  is, using higher value
>               than default ``1'' will allow running more than one
>               TA for the jobs at the thread.
> 
>               Do  note  that running more than one TA in parallel
>               may also require lowering  OVERFEED  value.   (E.g.
>               having a queue of 30 messages will not benefit from
>               more TAs, unless they all get something to process.
>               Having OVERFEED per default at 150 will essentially
>               feed whole queue to one TA, others are not  getting
>               any.)

Thats my point, I found out quite on accident this value means parallel
delivers per host/domain, it's never said anywhere.  Not in the SGML
beastie, the manpages, nowhere.  Near as I could/can tell, this value
controls the max per destination concurrency, and really only marginally
relates to the maximum number of threads....

I don't see any explanation of ZMailers concept of a thread, a job, and
a process, thats the main problem here.  And obviously fromt he way this
maxthreads value behaves it is different.  I would expect the definition
to be something more concise like, 'This controls the maximum number of
threads running to deliver to a single endpoint or channel/host pair at
any given moment.'

Thats what it appears to do, and one is *way* not enough for sites like
AOL.com which take 2 seconds (minimal) per message, this means that
1,200 messages take atleast 2,400 seconds.  There's also MSN.Com which
often times spins on a message for 20 seconds before accepting it.  And
these are big sites.

:)  So yes in practice it probably should be more than one if that is
indeed what it does.

It is the lates (err yah) 2.99.55 (or am I wrong this early inthe
morning, well I know its da.da.da.55).

Again, thanks much.

Michael Loftis

> 
> > Any help?
> 
> --
> /Matti Aarnio   <mea@nic.funet.fi>
-
To unsubscribe from this list: send the line "unsubscribe zmailer" in
the body of a message to majordomo@nic.funet.fi