Re: [pcp] I-D Action: draft-vinapamula-flow-ha-03.txt

ARP is applied at the radio bearer level which often represents multiple IP flows.

A user who is congested because their demand exceeds the radio resources they can receive, e.g. coverage challenge, from the network may never see ARP come into play if the radio network is not itself congested.

-----Original Message-----
From: pcp [mailto:pcp-bounces@ietf.org] On Behalf Of Tirumaleswar Reddy (tireddy)
Sent: Friday, September 19, 2014 5:46 AM
To: mohamed.boucadair@orange.com; pcp@ietf.org; cheshire@apple.com
Subject: Re: [pcp] I-D Action: draft-vinapamula-flow-ha-03.txt

> -----Original Message-----
> From: mohamed.boucadair@orange.com
> [mailto:mohamed.boucadair@orange.com]
> Sent: Friday, September 19, 2014 2:49 PM
> To: Tirumaleswar Reddy (tireddy); pcp@ietf.org; cheshire@apple.com
> Subject: RE: I-D Action: draft-vinapamula-flow-ha-03.txt
>
> Re-,
>
> Please see inline.
>
> Cheers,
> Med
>
> >-----Message d'origine-----
> >De : Tirumaleswar Reddy (tireddy) [mailto:tireddy@cisco.com] Envoyé :
> >vendredi 19 septembre 2014 11:07 À : BOUCADAIR Mohamed IMT/OLN;
> >pcp@ietf.org; cheshire@apple.com Objet : RE: I-D Action:
> >draft-vinapamula-flow-ha-03.txt
> >
> >> -----Original Message-----
> >> From: mohamed.boucadair@orange.com
> >> [mailto:mohamed.boucadair@orange.com]
> >> Sent: Friday, September 19, 2014 11:38 AM
> >> To: Tirumaleswar Reddy (tireddy); pcp@ietf.org; cheshire@apple.com
> >> Subject: RE: I-D Action: draft-vinapamula-flow-ha-03.txt
> >>
> >> Hi Tiru,
> >>
> >> Please see inline.
> >>
> >> Cheers,
> >> Med
> >>
> >> >-----Message d'origine-----
> >> >De : Tirumaleswar Reddy (tireddy) [mailto:tireddy@cisco.com] Envoyé :
> >> >jeudi 18 septembre 2014 17:18 À : BOUCADAIR Mohamed IMT/OLN;
> >> >pcp@ietf.org; cheshire@apple.com Objet : RE: I-D Action:
> >> >draft-vinapamula-flow-ha-03.txt
> >> >
> >> >
> >> >I can think of two to three ways to address this problem
> >> >1)Applications with required authorization can only signal that
> >> >it's a high priority flow.
> >> >2)End User with the help of a web-portal let's say provided by
> >> >Home Gateway or ISP can re-arrange priorities of flows.
> >> >Applications like uTorrent already provide UI so that user can
> >> >assign priorities for different file downloads.
> >> >3)Just like various good background applications use LEDBAT to
> >> >limit congestion, we can except that such applications in future
> >> >will explicitly signal that it's a low priority flow.
> >> >
> >>
> >> [Med] Do you mean PCP messages will be modified in path?
> >
> >No, I meant there will be some applications that will declare that
> >they have low-priority flows just like applications that declare
> >that HA is not required.
>
> [Med] OK. If we assume a priority level was added, what will be the
> behavior of the server when it receives a request for high-priority
> flow but there is no resources to honor the request? Should it remove
> entries with low-priority from the check-pointing table? I'm afraid
> this will induce more complexity at the server side.

I thought servers already have such complexity, for example Mobile Networks already use techniques like ARP (Allocation and Retention priority) to accommodate high-priority flow by penalizing low-priority flows.

>
> >
> >>
> >> >>
> >> >> Wouldn't (1) be sufficient to address the case you mentioned in
> >> >> your message?
> >> >
> >> >I don't think it is sufficient, it does not provide relative priority.
> >>
> >> [Med] Do these flows require check-pointing too? Or you only care
> >> about prioritizing these flows during congestion events?
> >
> >These flows could care about both check-pointing and prioritization
> >or care about only one of them.
>
> [Med] I can understand the priority level for flows that are subject
> to check- pointing, but I have an issue to generalize this draft to
> address the prioritization purpose ** only **. The problems are not
> the same. If prioritization is needed as a requirement in its own
> (which I understand since I'm for a mean to indicate at least DSCP),
> then a dedicated option can be defined for that purpose.

Works for me.

-Tiru

_______________________________________________
pcp mailing list
pcp@ietf.org
https://www.ietf.org/mailman/listinfo/pcp

________________________________

This e-mail may contain Sprint proprietary information intended for the sole use of the recipient(s). Any use by others is prohibited. If you are not the intended recipient, please contact the sender and delete all copies of the message.