<pre>
From glen.turner at adelaide.edu.au Sun Jan 31 13:43:14 1999
Return-Path: <glen.turner@adelaide.edu.au>
Received: from teatree.itd.adelaide.edu.au (root@teatree.itd.adelaide.edu.au 129.127.134.20)
by ssadler.phy.bnl.gov (8.8.7/8.8.7) with ESMTP id NAA04981
for <adler@ssadler.phy.bnl.gov>; Sun, 31 Jan 1999 13:43:11 -0500
Received: from adelaide.edu.au (gturner@teatree.itd.adelaide.edu.au 129.127.134.20)
by teatree.itd.adelaide.edu.au (8.8.7/8.8.7) with ESMTP id FAA04144;
Mon, 1 Feb 1999 05:13:19 +1030
Sender: gturner at teatree.itd.adelaide.edu.au
Message-ID: <36B4A447.BD4DF10@adelaide.edu.au>
Date: Mon, 01 Feb 1999 05:13:19 +1030
From: Glen Turner <glen.turner@adelaide.edu.au>
Organization: IT Services, University of Adelaide, South Australia
X-Mailer: Mozilla 4.08 en (X11; I; Linux 2.0.36 i686)
MIME-Version: 1.0
To: Stephen Adler <adler@ssadler.phy.bnl.gov>,
Micah Alpern <malpern@Princeton.EDU>
Subject: Re: Software Summary and Concerns about the future of the Net
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Status: OR
Micah Alpern wrote:
> When I first heard about this concept of prioritized data routing it
> seemed fine (obviously someone's ftping important research data
> should get higher priority than some freshman who's playing Quake in
> his room) but it is a very slippery slope we are sliding down here.
> The Internet II project is offering us sweets in one hand (Much more
> bandwith) and a muzzle in the other (packet prioritization). When
> (this is not a question of if, but when) Internet II technologies
> transitions into the commercial Internet the poor and unfortunate
> will be relegated to poorly paved secondary roads while the rich
> will keep the "Information Superhighway" for themselves.
Hi,
Micah's comments confuse the two aspects of differentiated network
services: quality of service (QoS) and class of service (CoS).
Part of the confusion is because QoS and CoS are delivered using the
same "differentiated services" technology. The technology basically
consists of preferential packet queuing (by looking at the precedence
bits in the IP header), some policing to stop the precedence bits
being faked, and some bandwidth reservation.
The QoS aspect is that some applications simply don't work unless the
network meets a particular level of service. The student playing
Quake *should* get a better QoS than someone FTPing data, as FTP
packets can be delayed with little user angst, whereas Quake is an
interactive application and excessive delays lead to your character
being blown away. Similarly, voice and video traffic should get
better QoS than the Quake player.
It is the QoS feature that Internet2 seeks to provide — not so much
for Quake players, but for video and supercomputer-driven
visualisation.
The second aspect of differentiated services is class of service
(CoS). Commercial ISPs are seeking to differentiate their customer
bases by offering differing classes of service: say First, Business
and Economy. This concept isn't new, as anyone who has booked an air
ticket will know.
As much as I dislike flying economy, I doubt it presents a threat to
my civil liberties. My view is that if Economy counldn't reach a
destination that First could reach, then there would be a civil
liberties problem. But that is not what is being proposed.
However, if Micah has a case to make, it is against classes of service
that the case should be made against, and not against quality of
service.
Commerical ISPs that are not also telcos are also interested in QoS.
They want provide voice-over-IP services and get a cut of the telco's
voice business. QoS is needed for VoIP as the packets have to be
delivered moments after a person speaks.
Because of these commercial pressures, ISPs are now planning to deploy
the differentiated services technologies required for CoS and QoS,
develop technology for ISPs regardless of Internet2 (which has well less
than a dozen gigabit routers).
Finally, a lot of networks currently provide QoS. All good routers
prefer BGP packets and penalise ICMP packets, just to promote the
stability of the network. A lot of non-US networks currently
prioritise traffic to give good performance when links to the US
congest. Typically USENET and e-mail traffic is given a lower
priority, as a few moments delay in the delivery of this traffic
hardly matters.
Very few networks currently provide CoS.
I hope this is helpful,
Glen
--
Glen Turner (08) 8303 3936 glen.turner at adelaide.edu.au
50% Network support specialist at the University of Adelaide
50% Australian Academic and Reseach Network technical guru
50% CompSci PhD dissertation writer
50% Linux tinkerer
----
200% stereotypical geek
</pre>
From glen.turner at adelaide.edu.au Sun Jan 31 13:43:14 1999
Return-Path: <glen.turner@adelaide.edu.au>
Received: from teatree.itd.adelaide.edu.au (root@teatree.itd.adelaide.edu.au 129.127.134.20)
by ssadler.phy.bnl.gov (8.8.7/8.8.7) with ESMTP id NAA04981
for <adler@ssadler.phy.bnl.gov>; Sun, 31 Jan 1999 13:43:11 -0500
Received: from adelaide.edu.au (gturner@teatree.itd.adelaide.edu.au 129.127.134.20)
by teatree.itd.adelaide.edu.au (8.8.7/8.8.7) with ESMTP id FAA04144;
Mon, 1 Feb 1999 05:13:19 +1030
Sender: gturner at teatree.itd.adelaide.edu.au
Message-ID: <36B4A447.BD4DF10@adelaide.edu.au>
Date: Mon, 01 Feb 1999 05:13:19 +1030
From: Glen Turner <glen.turner@adelaide.edu.au>
Organization: IT Services, University of Adelaide, South Australia
X-Mailer: Mozilla 4.08 en (X11; I; Linux 2.0.36 i686)
MIME-Version: 1.0
To: Stephen Adler <adler@ssadler.phy.bnl.gov>,
Micah Alpern <malpern@Princeton.EDU>
Subject: Re: Software Summary and Concerns about the future of the Net
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Status: OR
Micah Alpern wrote:
> When I first heard about this concept of prioritized data routing it
> seemed fine (obviously someone's ftping important research data
> should get higher priority than some freshman who's playing Quake in
> his room) but it is a very slippery slope we are sliding down here.
> The Internet II project is offering us sweets in one hand (Much more
> bandwith) and a muzzle in the other (packet prioritization). When
> (this is not a question of if, but when) Internet II technologies
> transitions into the commercial Internet the poor and unfortunate
> will be relegated to poorly paved secondary roads while the rich
> will keep the "Information Superhighway" for themselves.
Hi,
Micah's comments confuse the two aspects of differentiated network
services: quality of service (QoS) and class of service (CoS).
Part of the confusion is because QoS and CoS are delivered using the
same "differentiated services" technology. The technology basically
consists of preferential packet queuing (by looking at the precedence
bits in the IP header), some policing to stop the precedence bits
being faked, and some bandwidth reservation.
The QoS aspect is that some applications simply don't work unless the
network meets a particular level of service. The student playing
Quake *should* get a better QoS than someone FTPing data, as FTP
packets can be delayed with little user angst, whereas Quake is an
interactive application and excessive delays lead to your character
being blown away. Similarly, voice and video traffic should get
better QoS than the Quake player.
It is the QoS feature that Internet2 seeks to provide — not so much
for Quake players, but for video and supercomputer-driven
visualisation.
The second aspect of differentiated services is class of service
(CoS). Commercial ISPs are seeking to differentiate their customer
bases by offering differing classes of service: say First, Business
and Economy. This concept isn't new, as anyone who has booked an air
ticket will know.
As much as I dislike flying economy, I doubt it presents a threat to
my civil liberties. My view is that if Economy counldn't reach a
destination that First could reach, then there would be a civil
liberties problem. But that is not what is being proposed.
However, if Micah has a case to make, it is against classes of service
that the case should be made against, and not against quality of
service.
Commerical ISPs that are not also telcos are also interested in QoS.
They want provide voice-over-IP services and get a cut of the telco's
voice business. QoS is needed for VoIP as the packets have to be
delivered moments after a person speaks.
Because of these commercial pressures, ISPs are now planning to deploy
the differentiated services technologies required for CoS and QoS,
- irrespective* of Internet2's usage of the same differentiated
develop technology for ISPs regardless of Internet2 (which has well less
than a dozen gigabit routers).
Finally, a lot of networks currently provide QoS. All good routers
prefer BGP packets and penalise ICMP packets, just to promote the
stability of the network. A lot of non-US networks currently
prioritise traffic to give good performance when links to the US
congest. Typically USENET and e-mail traffic is given a lower
priority, as a few moments delay in the delivery of this traffic
hardly matters.
Very few networks currently provide CoS.
I hope this is helpful,
Glen
--
Glen Turner (08) 8303 3936 glen.turner at adelaide.edu.au
50% Network support specialist at the University of Adelaide
50% Australian Academic and Reseach Network technical guru
50% CompSci PhD dissertation writer
50% Linux tinkerer
----
200% stereotypical geek
</pre>