Cell Relay Archive

[Tony Lee <tslee@mail.arc.nasa.gov>]

Hi!

Some people like to use QoS and CoS interchangeably while others will
make a distinction. I'm one of the latter so here goes. CoS (or more
commonly Differentiated CoS) is used quite often to describe an architecture
which provides some limited QoS capability. The important characteristic
here is that it provides for the classification of traffic based on
some aspect (typically IP TOS bits, port number, etc), and different
traffic policies based on that classification. The simple example is
that each packet/cell would have a priority bit, and routers/switches
would behave differently based on that setting. I believe that
in the IP world 802.1p/q and IPv6 will use this mechanism.

This allows the network to prioritize certain traffic but does not
provide a mechanism for resource allocation and guarantees. The above
CoS architecture would work great if only my traffic had the priority
bit set, but will degrade as more high-priority traffic is added. The
point being is that while my traffic will be treated preferentially,
it is all relative to other traffic.

A full QoS system builds on this to add a more deterministic behavior
to the network. Basically it adds resource allocation which will
guarantee (barring physical layer failures) that the network will
provide my traffic with a certain level of service. Two examples
of this are RSVP and PNNI/UNI.

I have seen CoS described as 'soft QoS' where the 'soft' aspect comes
from the lack of guarantee. I tend to believe that this will be
sufficient for 95% of networks and users, and certainly comes at
a much lower price than a full QoS architecture. At this time it appears
that it may be the only viable solution for IP and the Internet and has
been recently demonstrated by Lawrence Berkeley Labs in a video
connection across the Internet to Argonne Labs in Illinois. The
press release is at:
http://www.lbl.gov/Science-Articles/Archive/internet-priority-traffic.html

However, that last 5% of mission critical apps will still require a QoS
infrastructure such as PNNI/UNI and ATM.

Hope this helped,

-tony


Tony Lee
WAN group, NAS division
NASA Ames Research Center

tslee@mail.arc.nasa.gov

<Standard Disclaimer applies>



At 1:43 PM +0200 4/27/98, Mike Vogeleer wrote:
>Hi,
>
>Can anyone explain me (in general + some simple examples) the difference
>between QoS and CoS (class of service). Is the term 'CoS' ever really
>used?
>
>Many thanks,
>Mike.