The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Sep> msg00395



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Any SPs using QoS ???

  • From: Sham Chakravorty <schakra@mitre.org>
  • Date: Wed, 27 Sep 2000 13:20:48 -0400
  • CC: Panos Trimintzios <p.trimintzios@eim.surrey.ac.uk>, Lixia Zhang <lixia@cs.ucla.edu>, Sudheer Dharanikota <sudheer@nayna.com>, Martin Picard <mpicard@sinc.ca>, mpls@UU.NET
  • Organization: The MITRE Corporation

Graham,

Excellent points.  I agree; just because BW is commodity (a market
started by Enron - a power trading/building co.),  it a'nt cheap.  Try
getting a T3-size BW for a week or a day.  Next, I like to argue you
will still need QoS - which is really prioritizing traffic.  If the
roads were as wide as a field (BW glut), it doesn't mean there ought not
to be some form of taffic control.  You will always need to let the
ambulance go before all the Toyotas (flow-based control => IP switching?
does not work, re Ipsilon, meaning larger users of BW getting all the
preference is not much good unless of course you ckt-switch [or
lambda-switch] them based on SLAs). 

Also, you will never have enough BW in the wireless media and possibly
in the last mile; not to mention we have not really put any voice or
video on the Internet (or enterprise networks) to speak about.  When
they come, watch out.  This BW commodity could be as pricy as "gold" and
major controls (of all kinds!) will be needed.

Finally, the mechanisms for QoS should be network-imbedded and not
necessarily be an operational issue on an on-going basis (for any SP or
enterprise).  See the attempts Cisco is making with its CDN concept
along with Cable and Wireless sort of a backdoor approach (VPN based?).
This means it is time for a new, robust, QoS technology.  Unfortunatly,
some of us believe that MPLS is not it.  After nearly four years of work
by hundreds(?) of people, we still don't have the framework for a
network-imbedded QoS concept or implementation. We still need ATM to run
MPLS for CBR, don't we? 

Chakravorty
***************************************************************

Graham Cope wrote:
> 
> Panos Trimintzios wrote:
> >
> > Lixia is right bandwidth is officially a market commodity see:
> >
> > Dow Jones to launch bandwidth index
> > http://home.cnet.com/category/0-1004-200-1677726.html
> >
> 
> Being a commodity, however, should not be confused with being cheap.
> There are also commodity markets in diamonds, gold and crude oil
> (topical at the moment!).
> 
> That then leads back to the fundamental (and possibly off topic)
> question as to whether IP QoS is really required.
>   To do so a service provider must charge for the good QoS. Can they
> really be bothered to put in such mechanisms? Where does that leave the
> cosy transit agreements that ISPs traditionally have?
>  If they cannot be bothered to charge (i.e. it costs more than they get
> back from it) then the Internet will remain best effort. I suspect the
> IP QoS will remain a niche opportunity for suppliers of premium class
> business VPNs with over-provisioning being the long term solution for
> most customers as bandwdith becomes cheap (it may be a commodity, but it
> ain't negligibly cheap yet, especially in the access network).
> 
> Just a thought.
> 
> Graham