PWS headless - Lack of Responsiveless and Double Submits

Peter William Lount peter at smalltalk.org
Thu Aug 26 00:31:36 UTC 1999


Hi,

http://www2.smalltalk.org:8080/chat

By putting the code, (Delay forMilliseconds: 100) wait, back into the
PWS>loopOnPort:loggingTo: method by and varying the milliseconds down to 1
i've gotten better behavior out of the headless squeam pws vm. It now
chews up just a few percent or even less than 1 percent of cpu power. It
also seems to go to sleep and chew up no cpu power.

Can someone please explain what is happening in this PWS
loopOnPort:loggingTo: method and how it interacts with the Squeak
"simulated" processes...

On ocassion the pws still doesn't respond. Squeak looks like it's gone to
sleep with 0.05% or less cpu usage... any explainations?

When processing the request it seems to chew up just under 5% of the cpu
power (on a 300Mhz Intel Pentium 2 processor). This means, with enough ram,
that there could be at least 20 copies of squeak running giving instant
response per cpu. Two cpus so 40 squeaks! I'll test this some time this
week to see actual performance... but I don't have enough ram to hold 40
squeak in ram....oh well...

Is there a way to have Smalltalk level code read the virtual machine unix
command line arguments... I'd like to have squeak read an optional argument
for the port that it should serve pages on... This allows many copies of
the same vm to be started simply via a single unix script...

All the best,

Peter William Lount
peter at smalltalk.org
http://www.smalltalk.org

----------
From: Peter William Lount <peter at smalltalk.org>
To: Squeak Mail List <squeak at cs.uiuc.edu>
Cc: Kellie Miller <kellie-miller at yahoo.com>; Norton, Chris
<chrisn at Kronos.com>; Tim Rowledge <rowledge at interval.com>; Mark A. Schwenk
<mas at wellthot.com>; Bert Freudenberg <bert at isgnw.CS.Uni-Magdeburg.De>
Subject: Re: PWS headless - Lack of Responsiveless and Double Submits
Date: August 25, 1999 4:31 PM

Hi,

http://www2.smalltalk.org:8080/chat

Well I've got it responding instantly! However it's chewing up 99.02% of
one of the two cpu's in the server box!!! As a test, and to try to
understand what is happening in the PWS main loop process I removed the
"(Delay forMilliseconds: 100) wait" from the PWS>loopOnPort:loggingTo:"
method. Response time went from delays to instant response. However the
darn loop is polling for packet connections on the ServerPort....  Can
someone explain what is happening in this loop and what the rational for
the Delay is? If it's designed to be polling that would be a poor design
choice... since the polling chews up all the cpu power...

Is there any way to have Squeak respond to packet arrivals with an
interrupt or process waking up?

I'd like to create a Apache Module if none exists, modsqueak, for Apache to
talk with squeak vm processes via a tcp packet or shared memory. However,
it would still require squeak to be able to respond instantly to a incoming
packet without having to poll for the darn packet... 

All the best,

Peter William Lount
peter at smalltalk.org
http://www.smalltalk.org

----------
From: Peter William Lount <peter at smalltalk.org>
To: Squeak Mail List <squeak at cs.uiuc.edu>
Cc: Kellie Miller <kellie-miller at yahoo.com>; Norton, Chris
<chrisn at Kronos.com>; Tim Rowledge <rowledge at interval.com>; Mark A. Schwenk
<mas at wellthot.com>; Bert Freudenberg <bert at isgnw.CS.Uni-Magdeburg.De>
Subject: PWS headless - Lack of Responsiveless and Double Submits
Date: August 25, 1999 3:56 PM

Hi,

Well the headless server now seems to responding most of the time. It still
doesn't respond all of the time when a request is made from a web browser.
Anythoughts?

In addition, once the page does respond subsequent requests form the "post
your message" submit button just hang it unless you press the button twice
in which case it responds instantly but with a duplicate entry (since it
was pressed twice!)... ;--(

You can try it (if it doesn't work the first time try again and make sure
to try double clicking on the "post your message" button) at this demo site
http://www.zoku.net:8080/chat

It's very strange in that it consistently won't respond the first time (or
two) but
if you try again time it's there instantly! This applies for when you
activate the page from a browser and when you press the submit button!

Anyone got any ideas about this?

All the best,

Peter William Lount
peter at smalltalk.org
http://www.smalltalk.org





More information about the Squeak-dev mailing list