I am cleaning out some old email and just read this. 

I also do not think that 90% of the TCP traffic goes thru mainframes, but I 
have a few different thoughts on the thoughts found below. 

First I am not sure what folks want to think of as mainframes. Some folks only 
think of what was called  os, dos, vse, mvs, z/os..., the group of IBM 
mainframe batch operating systems. 
Other might have counted the bunch or Univac, Dec, Honeywell, etc. 
And in today's world should a Sun 12K or an windows data center machine count 
as a mainframe?
How about large clusters?


I do remember mainframe p2p apps many years ago. Things such as bitnet for 
sharing email and files was a heavy user of network  bandwidth in my last job 
in the early 90's. This included our DEC Vax cluster and the IBM VM mainframe. 

I believe that Apache software is at the core of IBM's current web application 
server software. And this can and is run from mainframes. But I do not have any 
back of the hand idea of the percent of usage of the internet bandwidth. 

Also I wonder about the amount of traffic running on mainframes for 
applications such as banking or large retail online sales applications. Of 
course as Aaron pointed out that this traffic may use both the mainframe and a 
*nix system to get the job done. 

Another thing to wonder about is the use of the network using some form of vpn 
instead of private phone circuits. I assume that a portion of this is for 
mainframe networking. But how much is mainframe based vs *nix vs windows I do 
not know. 

I suspect that the lawyers at organizations such as Aaron's and the large phone 
companies have some estimation of the numbers for their current battles in 
Washington over changing the current charging models for network. 

len

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Aaron S. Joyner
Sent: Friday, October 06, 2006 3:55 AM
To: Triangle Linux Users Group discussion list
Subject: Re: [TriLUG] The thin line...

Andrew Ball wrote:

> ...Also, over 90% of TCP traffic goes through mainframes...

I think it's reasonably safe to call the bluff on that statistic.  
Probably well upwards of 30% of all TCP traffic is HTTP these days*, and 
last time I checked Apache accounts for around 62.52% of all web 
servers, Microsoft products another 30.13%, together around 93%**.  Just 
from looking at the rough numbers there, even if I'm high balling the 
HTTP numbers (probably not by much, if at all), then that's 25%+ that 
isn't touching a main frame at all.  The traffic goes straight from a 
client, likely through some appliance NAT router of theirs, into a cable 
or DSL modem, through an access device at the other end onto Ethernet, 
through a crap-load of routers, and eventually to a switch and some 
random unix/ms server*** on the other end, and back again.  Another 
interesting argument is that I've seen a lot of reports lately that p2p 
traffic is consuming crazy percentages of all internet traffic these 
days, some say as high as 60 to 75%.  I think those are exagerated 
numbers, but I'd believe 40% or more.  I am not even aware of any p2p 
apps for OS360, or anything even resembling a mainframe.  :)

Aaron S. Joyner

* - Alert: Conservative wild flaming guess.  Couldn't easily find any 
trustworthy statistics here.
** - http://news.netcraft.com/archives/web_server_survey.html
*** - Don't even try the "these are virtual instances on z-series 
mainframes" angle, no way there's any sizable percentage of websites on 
such a platform, see Netcraft again.
-- 
TriLUG mailing list        : http://www.trilug.org/mailman/listinfo/trilug
TriLUG Organizational FAQ  : http://trilug.org/faq/
TriLUG Member Services FAQ : http://members.trilug.org/services_faq/
--
TriLUG mailing list        : http://www.trilug.org/mailman/listinfo/trilug
TriLUG Organizational FAQ  : http://trilug.org/faq/
TriLUG Member Services FAQ : http://members.trilug.org/services_faq/

Reply via email to