On Thu, 16 Feb 1995 00:03:50 EST Bill Gruber said:
>On Wed, 15 Feb 1995 16:39:23 GMT Samer Farha said:
>
>>Actually, LISTSERV is one of the most complicated and hence CPU intensive
>>programs out there.
>
>I really have to disagree with that statement.  Yes, LISTSERV *is* rather
>complex, but it is also very efficient at what it does.  In my experience
>it doesn't use very much of the CPU at all - I mean well under 1%.
>
>Of course, it generates mail, and some of that mail uses other resources on
>its way out to the world, such as passing through mailer, smtp etc.  And in
>my opinion, if you want to talk about junk, talk about IBM's SMTP.
 
You said it!
 
     And here are some real number to prove it.  During this time
frame, SMTP delivered 10.8 million pieces, many of which were also
touched by MAILER and LISTSERV (at a fraction of the cost).  RSCS,
during the same time span, processed 3.8 million.
 
        LISTSERV@UBVM Volume Statistics for 199501
 
                                  Month          Daily
                                  Total        Averages
=======================================================
Postings to mailing lists:        28152          908.13
Recipients:                     9381405       302625.97
 
Digests issued:                    1439           46.42
Recipients:                       68082         2196.19
Indexes issued:                     738           23.81
Recipients:                        2400           77.42
 
DISTRIBUTE jobs processed:       438116        14132.77
Internally generated:             58481         1886.48
 
Outbound DISTRIBUTE jobs:       1006658        32472.84
Outbound NJE files:                3834          123.68
Outbound files to MAILER:        521657        16827.65
In non-BSMTP format:             125077         4034.74
Recipients:                    10960243       353556.23
 
GLX requests:                      1686           54.39
 
LISTSERV 1.8a with FIX18AX2 (48M)
  USERID: LISTSERV  DATE: 01/01/95 - 01/31/95
       Connect Time      :       741:08:41    hh:mm:ss
       Virtual CPU       :       895:38.80    mm:ss.ss
       Overhead CPU      :       459:45.88    mm:ss.ss
       Paging            :      10,960,131    Pages
       Spooled I/O       :     111,651,345    Records
       Non-Spooled I/O   :       7,996,040    SIOs
 
RSCS 3.1.1 RSU9403
  USERID: RSCS      DATE: 01/01/95 - 01/31/95
       Connect Time      :       741:04:46    hh:mm:ss
       Virtual CPU       :       950:07.47    mm:ss.ss
       Overhead CPU      :      1469:08.95    mm:ss.ss
       Paging            :      30,280,372    Pages
       Spooled I/O       :     110,447,895    Records
       Non-Spooled I/O   :         428,411    SIOs
 
LMAIL 1.2a
  USERID: MAILER    DATE: 01/01/95 - 01/31/95
       Connect Time      :       741:08:25    hh:mm:ss
       Virtual CPU       :       227:00.81    mm:ss.ss
       Overhead CPU      :       402:14.83    mm:ss.ss
       Paging            :      14,292,112    Pages
       Spooled I/O       :     194,699,595    Records
       Non-Spooled I/O   :           6,939    SIOs
 
SMTP 2.2
  USERID: SMTP*     DATE: 01/01/95 - 01/31/95
       Connect Time      :      6520:48:43    hh:mm:ss
       Virtual CPU       :      7865:21.47    mm:ss.ss
       Overhead CPU      :      2960:25.06    mm:ss.ss
       Paging            :      11,377,703    Pages
       Spooled I/O       :     119,670,200    Records
       Non-Spooled I/O   :     355,524,422    SIOs
 
A lot of my problems would go away if IBM could achieve even a modest 50%
increase in performance for SMTP, giving me back 90 hours of CPU time, and
this is on a 3090-300J!
 
I am currently installing/testing TCPIP 2.3, but I do not think there are any
improvements in this area.
 
 
Steve
([log in to unmask] | [log in to unmask] | (716)645-3564)