Re: [manet] Highest priority flow in DiffServ model

[Marshall Eubanks <marshall.eubanks at gmail.com>]

Hello;

On 8/28/05, Roy, Radhika (AEAD) <RoyR at saic-abingdon.com> wrote:
>  
>   
>  
> H, Susan: 
>   
> Please feel free to continue your explanations related to QOS as I am
> working on Army's 21-century worldwide network whose MANET-based networking
> is the key foundation. 
>   
> Also, I am in the middle of M&S works of the MANET-based network and will
> continue participating in technical discussions especially related to QOS as
> far as practicable publicly. 
>   
> Definitely, as you have put, QOS is about some calls will get what they are
> supposed to get related to bandwidth and other resources, while others will
> get less. 
>   
> The other thing is that the topology is changing in MANET networks.
> MANET-based routing protocols will have the layer-3 topology information for
> example. So, here are some probable scenarios: 
>   
> 1. Applications will express their QOS needs at the time of the call set-up
> (e.g., SIP). 
>   
> 2. Servers like QOS policy/CAC admission need to get involved to bridge the
> inter-layer information related to QOS and topology (application layer-to-IP
> layer) to indicate what needs to be done related to QOS in end-to-end path. 
>   
> 3. Servers like Bandwidth Broker/CAC need to signal IP routers what need to
> be done across the network including DiffServ and RSVP. (Simplified model
> from top to down - instead of end-to-end.) 
>   
> 4. If the MANET network changes topology after establishment of the calls,
> QOS policy/BB/CAC servers need to take care of QOS with changes in topology
> without involving call layer's call control servers. It is expected that
> this situation may cause some problems initially for the real-time traffic
> for delays in setting up the new paths, but it will be OK eventually. 
>   
> 5. For multicast in MANET environment, there are a lot of other things need
> to be taken care of. 
>   
 
If I may ask, how important is multicast to your worldwide network ? Is this
something that you rate as very important for Manet ?

Regards
Marshall Eubanks


> Anyway, these are couple of observations as inputs to stimulate discussions.
>   
> Best regards, 
>   
> Radhika R. Roy 
> 
>  ________________________________
>  From: Susan Hares [mailto:skh at nexthop.com]
> Sent: Sun 8/28/2005 9:23 AM
> To: Roy, Radhika (AEAD); Frye, Robert J (US SSA); manet at ietf.org
> Subject: RE: [manet] Highest priority flow in DiffServ model
> 
>  
>  
> 
> Roy and Robert: 
> 
>   
> 
> Correctly put – QoS is about some devices getting less of the bandwidth. 
> 
>   
> 
> In 1999, internet-2 Bandwidth broker (BB) sub-group did some work on a BB to
> BB protocol. We experimented with TE based on OSPF, Radius/Diameter and
> RSVP-TE.  [Yes I am admitting I ran some of the early trials.]   Only a few
> routers (such as the Hitachi router) passed the early Diff-Serv field trials
> in 1999-2000 to provide high quality QoS based on the Diff-Serv routing. 
> 
> In the subsequent years, cisco and a host of routers have put the hardware
> and software in line to guarantee QoS. 
> 
>   
> 
> Hitachi put the effort in to work on the multi-media (Video, data, etc). 
> 
>   
> 
> In the 6 years since 1999, 
> 
> -        TWEG working group – has championed TE pipes over IP, 
> 
> -        MPLS has champion off-line CSPF for traffic engineered pathway
> selection, and the 
> 
> -        PCE work is trying communicate between a path-calculation box and
> the node. 
> 
>   
> 
> Other alternatives are using RSVP-A and RSVP-TE pathways that walk the path
> obtaining the information.  
> 
>   
> 
> So – there is a significant body of work that exists in the IETF and in
> research journals on traffic engineered path. 
> 
> I have hit only the highlights of the work in the last 7-10 years. 
> 
>   
> 
> The key issues around QoS for Manet are extremely interesting.  Not only do
> you have less bandwidth, but the delay issues due to varying convergence
> impact any delay sensitive traffic services. 
> 
>   
> 
> So – the key point is to utilize fast-convergence or pre-calculation of
> paths. 
> 
> The difficult point is bandwidth brokering multicast.  
> 
>   
> 
> So.. before I launch into more details.  Let me know if you are still
> interested. 
> 
>   
> 
> Sue Hares 
> 
>   
> _______________________________________________
> manet mailing list
> manet at ietf.org
> https://www1.ietf.org/mailman/listinfo/manet
> 
> 
>

_______________________________________________
manet mailing list
manet at ietf.org
https://www1.ietf.org/mailman/listinfo/manet