Great Circle Associates List-Managers
(August 1996)

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: Reply from Compuserve
From: Eric Thomas <ERIC @ VM . SE . LSOFT . COM>
Date: Wed, 7 Aug 1996 17:39:32 +0200
To: list-managers @ GreatCircle . COM
In-reply-to: Message of Tue, 6 Aug 1996 17:28:34 -0700 (PDT) from list-managers-owner @ GreatCircle . COM

On Tue, 6 Aug 1996  17:28:34 -0700 (PDT) Dave Bigham <dbigham @
 cantec .

>I think it is important to note that the above statistics are individual
>statistics for all of the 166,000+  messages sent from all of the lists.
>They are not from  one list at one time. So, for  example, *each one* of
>the fastest 50% took 8 seconds to send - each.

If you write down all the  166,000 individual delivery times from fastest
to  slowest,  the  median will  be  8  seconds.  That  is, 50%  of  these
deliveries will have taken  8 seconds or less, and 50%  will have taken 8
seconds or more.

>Thats more that  664,000 message-seconds, if that makes  any sense. It's
>not a case of 80,000+ messages being sent in 8 seconds or less.

You're right, but what the users see  and care about is the time it takes
them to get their copy of a message  posted to the list, ie the 8 seconds
or whatever, as  opposed to the 664,000.  I moderate one of  the lists on
the server in question  (I'm in Sweden, the server is in  the US), and it
usually takes 5-10  seconds between the time  I press the SEND  key on my
approval request and the time my copy  of the posting is in my mailbox. I
get this  copy because I'm subscribed  to the list and  the message isn't
from me, it isn't something that is sent separately before the mailing is
processed. The list has about 400 recipients, and the server delivers 500
messages  in  parallel,  so  most  of  the  time  my  host  is  contacted
immediately and it takes a few  seconds to deliver the message, hence the
result. For  a much  larger list I  might have to  wait behind  2-3 other
people which would receive service before me. In general, with very large
workloads the median increases  to 30 sec or so while  the 99% average is
not significantly affected (normally still under 2 minutes). While 30 sec
is not  quite as impressive, it  makes no difference to  the average user
who spends over a minute reading each message.


Indexed By Date Previous: Re: Reply from Compuserve
From: Eric Thomas <ERIC @ VM . SE . LSOFT . COM>
Next: Disabling "who" completely in Majordomo
From: Dean Ericksen <dean @ onenw . org>
Indexed By Thread Previous: Re: Reply from Compuserve
From: Eric Thomas <ERIC @ VM . SE . LSOFT . COM>
Next: Re: Reply from Compuserve
From: Brock Rozen <brozen @ netvoyage . net>

Search Internet Search