nanog mailing list archives

Re: Internet II is coming... (fwd)


From: Oded Comay <comay () post tau ac il>
Date: Mon, 14 Oct 1996 23:32:45 +0200 (IST)

My apologies for sending it to all of you. I am not sure who is
interested in the subject. 

Traffic prioritization as implemented by routers is built around an
interface output queue. In the case where a router is connecting a fast
LAN to a relatively slow line, the queue might get built on the slow line
interface (in the outgoing direction) but not the other way around. In the
even more special case (but probably the common one) where the traffic is
asymmetric so incoming traffic is significantly larger than outgoing one,
no queue will be created, and the prioritization scheme will not be
operated at all. 

Hope this helps,
Oded.

----- Forwarded message from Michael Ciavarella -----

---------- Forwarded message ----------
Date: Wed,  9 Oct 96 07:04:08 UTC
From: Hank Nussbacher <hank () ibm net il>
To: Sean Doran <smd () cesium clock org>, "Dorian R. Kim" <dorian () cic net>
Cc: nanog () merit edu
Subject: Re: Internet II is coming... 

.....

The reason for this is that within the university, access and use is unlimited
and uncontrolled.  The student and the professor have equal access from their
workstation.  That is why the Israeli university consortium has come up with
a different solution.  It is called chokepoint.  A unix system that acts
as a firewall/gateway.  If the total access speed to the Internet is 
T1 then at the chokepoint one can define that port80 can use a maximum of
700kb.  And one can define that telnet is guaranteed 30kb.  And that 
10.2.1.1 is guaranteed 128kb no matter what.  This way, faculty server and
faculty workstations can be given priority over student access.  In addition,
no one faculty member can "hog" the system.  Facultycan even pay to the
chokepoint to improve their service over others.  An entire set of rules
based on protocol and IP address can be set up and implemented.  They have
been running this way for the past 6 months.  


How does this differ from the queuing and prioritisation features on,
say, a cisco?  Obviously router load is reduced by moving this type of
"service" to a dedicated box, but unless you're using the UNIX box as
your router to the net, how does it exert this type of control over your
bandwidth?  Does it drop incoming packets which exceed a (set) threshold
even though they've already come over your link, or does it sit 
at your provider's end, in which case how do you control your outgoing
traffic?


mike.. (blah.  mornings.)

----- End of forwarded message from Michael Ciavarella -----




- - - - - - - - - - - - - - - - -


Current thread: