I have to agree with a lot of the comments that P Kavi made. My greatest
issue with any attempt at GOS is that one MUST be able to control ingress
traffic. Unless one can do so in a fairly clean manner, I don't see how it
will work in practice.
* No policing at ingress: You can't have QoS unless
you can limit how much traffic enters the network,
and discard, or at least mark the excess traffic.
* No effective Class of Service mechanism:
The support issue between ASes is, I believe, the largest one. I think
getting such a policy in place will dwarf the carrier agreement woes
between major Telcos. Can you imagine trying to build a common framework
for Qos/Cos across the majority of providers any time soon? The
administrative headaches will, I think, far outweigh any technical
issues...
Fortunately, we do have the advantage of hindsight and can attempt to
avoid the worst part of the telco industry's carrier agreements. It may
not be easy but it is an essential problem to solve if we want packet
switching IP networks to become the main transport mechanism for
information in the next century.
That's why we need lots of knobs in the routers, and the CPU power, etc.
to back them up. We don't need to convert the entire network in order to
implement usable realtime voice services; we just need to allow for a
portion of the bandwidth to be allocated according to QoS/CoS. There are
many ways this could be done without implementing metering, if the routers
at every point on the path, can control the use of QoS/CoS.
Michael Dillon - Internet & ISP Consulting
http://www.memra.com - E-mail: michael@memra.com
The bottom line is track record. Not track tearing. Not track derailing.
But pounding the damn dirt around the track with the rest of us worms.
-- Randy Bush