Re: Two off the wall questions

[Steve Williams <steve@genie96.com>]

> 
> 
> On a normal day, we send out around 2000 faxes.  On Some days, we sound out
> as much as 8000 faxes.
> 

WOW!!!

I would guess that these are small faxes, because according to my math,
7 modems & 24 hours & 60 minutes/hour = 10080 modem minutes, and divide
that by 8000 faxes, leaves an average time of 1.26 minutes per job, assuming
no overhead or transmission failures.

> When we get past that 2000 mark, Hylafax really starts to bog down.  It
> still sends OK, but submitting new faxes take longer and longer and longer.
> We have thrown money at hardware to make it work faster, but it still causes
> us grief when the queue gets full.  We think that the root of the problem,
> is that Hylafax does directory scans each time it needs information, instead
> of keeping a state file.
> 
> Anyway, to help eleviate the problem, I want to set up a seperate queue for
> our marketing people to use.  That way when they are sending out gobs of
> faxes, it won't effect the general user so much.

Humm,

When you say above that above the 2000 mark it bogs down, do you mean 2000
Queued faxes?

You say the queueing of the faxes takes a long time.  Is this the time for
the "sendfax" command to complete?  How big are the faxes that are being
submitted?  How long do you consider "bogging down". ie:  how long does it
take for the sendfax command to complete?

> 
> We have a dedicated system just for running Hylafax 4pl1:
> 
> AMD K6 200 MHz running RedHat Linux 4.1
> Comtrol Rocketport PCI serial card driving 7 multitech fax modems.
> 

It sounds like this HylaFAX is serving a mission critical function, so I
would imagine that you wouldn't want to be messing around with it too much.

>From your previous post, I would presume that you are a programmer if you
are willing to even consider delving into the HylaFAX source code.

There are some things that I would suggest prior to trying to divide up the
queue directories:

In my experience, when looking to optimize code, the first optomization 
is generally fairly easy to have a significant impact on performance.  After
that, it gets harder & harder to get the performance boost.

It might be one thing that is the bottleneck, and removing that bottleneck
might make the difference that you need.

How to accomplish this?

One way would be to compile up HylaFAX with profiling support enabled.
( -p flag to cc ).  I would HOPE this is supported by Linux.

This would enable you to analyze where the time is being spent in the 
program, be it 1000's of procedure calls, or maybe just one intense 
subroutine that takes forever.

The other thing I might consider to help narrow down where the problem down
would be to put a "timestamp" on the output produced by sendfax -vv, and
take a look at exactly in the process is taking the time.

( You may have done all this already ).

Anyway, I am intrigued by your problem, but probably not as much as you
are frustrated by it <g>.

I am moderately familar with the HylaFAX software, after debugging through
it to port it to the AIX platform.  Feel free to contact me with ideas/
questions.

> 
> On Fri, 13 Jun 1997, Steve Williams wrote:
> 
> > Hi,
> > 
> > There is no "easy" way to do it.  I can imagine that you would end
> > up beating yourself up over it. 
> > 
> > Maybe if you describe what you are trying to accomplish, there is another
> > way that it could be resolved.  There are many powerful features that
> > lots of people are not aware of.
> > 
> > Good Luck,
> > 
> 
> Thanks,
> 
> John
> --
> 
> |         |
> +--+  ==  |  John Patrick Poet          Blue Sky Tours
> |  |      |  System Admin/programmer    10832 Prospect Ave., N.E.
> |     +---+  john@BlueSkyTours.COM      Albuquerque, N.M. 87112
> |     |      Ph. 505 293 9462           Fx. 505 293 6902
> 
> 


-- 
	Steve Williams, Calgary, Alberta, Canada
	Genie Computer Systems Inc.
	steve@genie96.com