Hi Pascal,
Thank you for your comments on my article. In the example, /tmp/cgi42msg is
a standard bytestream file. I call it a pseudo message file because it
performs many of the same functions as an MPE message file. You are correct
in pointing out that the file grows over time. This doesn't create a
problem for me because each record is small and I shut everything down when
backing up the system.
The Posix shell script opens, appends to, and closes cgi42msg for each
transaction, creating the file if necessary. So, the listening program can
monitor the size of the file and do a rename and build when needed. If a
shell script creates the file first, the listening program's build will fail
and this can be ignored.
Frank J. Gribbin
Potter Anderson & Corroon LLP
1313 N. Market Street, 6th Floor
Wilmington, DE 19803
Phone: (302) 984-6155
Telecopier: (302) 658-1192
E-mail: [log in to unmask]
************************************************************************
This electronic mail transmission may contain confidential or proprietary
information intended only for the person(s) named. Any use, distribution,
copying, or disclosure by another person is strictly prohibited.
************************************************************************
-----Original Message-----
From: Pascal Luyckx [mailto:[log in to unmask]]
Sent: Tuesday, September 18, 2001 4:04 AM
To: 'Frank Gribbin'
Subject: RE: [HP3000-L] APACHE performance
Hi Frank,
I've read your article and I have a question about it.
What is a pseudo message file? Is it a simple flat
file or is it a special kind of message file? If it
is a flat file, and you open it in read only mode
in the listening program, the pseudo message file
grows over time, unless you empty it after each run
of your listener program. How do you handle this?
Regards,
Pascal Luyckx
Cheops technology nv
Jozef Verbovenlei 91
2100 Antwerp
Belgium
Phone: +32 3 366 50 30
Fax: +32 3 366 57 35
-----Original Message-----
From: Frank Gribbin [mailto:[log in to unmask]]
Sent: Saturday, September 15, 2001 7:45 PM
To: [log in to unmask]
Subject: Re: [HP3000-L] APACHE performance
Hi Joe,
Here is a link to an article I wrote on performance issues surrounding
Apache/iX CGI. I hope it helps.
http://www.interex.org/pubcontent/enterprise/jul01/04webix.jsp
Frank Gribbin
On Thu, 13 Sep 2001 11:20:35 -0700, Joe Weisman <[log in to unmask]> wrote:
>Well, I was attacking this problem a few months ago and never got it done,
>and now I am back.
>
>What I have is APACHE running on my HP3000, and my own data server also
>running on the HP3000. My server is an existing MPE process which supports
>lots of users
>at dumb terminals connected to the HP via serial lines, TCP/IP, or
>whatever. I need to support additional users running on web browsers
>connected to the internet, who will be doing the same transaction
>processing on the same data bases as the terminal users.
>
>For each 'study' running on my server, there is an HTML page on the HP
>which has a pushbutton to start a CGI process that will then open
>communication with my server, and exchange data with it, and send a new
>page to the browser with whatever the transaction processing script says to
>say. Then the browser user can answer questions on the page and the
>answers are sent back to my server and turned into another html page which
>is sent back to the browser, and so on.
>
>Performance is a serious issue, and the performance I am able to get so far
>is abysmal. I can handle the delays between the CGI process coming up and
>it contacting my server, and my server getting back to it, and all that.
>What is so far intractable is the start-up time for the CGI process. The
>best I can get is never less than five seconds, from 'click' on the button
>to any response from the CGI process.
>
>My simplest test, where the original HTML page is basically only the
>button, and the script it executes is just an echo of a couple environment
>variables, the response is still never less than four seconds. Same delay
>if the script uses 'callci' to do a PRINTOP; I push the button and several
>seconds later I see something on the operator console.
>
>I hope (hope hope) I am doing something ignorante and/or stupid which can
>be fixed, and that someone out there, or the HP response center, or some
>yet unlocated paid consultant, can help me out here.
>
>The URL for my trivial test is http://hpspec.cfmc.com/cgi-bin/echotest.test
>Be well, and do good things.
>
>- Joe Weisman
> Computers for Marketing Corporation
> 541-745-5265 (voice)
> 541-745-5266 (fax)
> [log in to unmask] (e-mail)
>
>* To join/leave the list, search archives, change list settings, *
>* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
