[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MEXT] weird idea about flow bindings

George Tsirtsis escribió:

I am not sure I fully understand what you propose.

If I understand correctly your scheme seems to assume the following:
1) what constitutes a "flow" is something fixed and is understood by both the MN and the HA. 

i don't think i am assuming that
I assume that the initiator of a flow will set the flow label and keep it unchanged for the duration of the flow
2) "flows" start one at a time giving enough time to the MN, to send a BU to the HA to label (and optionally redirect) each one, before the next starts.
no, i am assuming that the flow starts and upon the reception of the first packet of the flow, the MN redirects it to the correct CoA using a BU message. Basically, i am assuming that most of the flows that i care about are longer than one RTT between the MN and the HA. Is this safe? 

Regards, marcelo
These are far from safe assumptions, but maybe I did not understand your proposal in the first place. 

George
On Sat, May 16, 2009 at 7:47 PM, marcelo bagnulo braun <marcelo at it.uc3m.es <mailto:marcelo at it.uc3m.es>> wrote: 

    Hi,

    Here goes a weird idea about how to do the flow bindings.
    What about if we only identify flows using the flow label?
    I mean, that instead of having to define a format to describe the
    flows, we simply include the flow label value of the specific flow
    in the flow bindings rules.

    This would imply basically that the MN host has the policy
    configured about which flow will flow through each CoA.
    the first packet of each flow will be received through the primary
CoA (the one with highest priority). 

    Upon the reception of the flow, if the MN wants to divert the flow
    through an alternative CoA,  it sends a BU with the Flow binding
    option, including the Flow label value of the flow and the BID
    associated to the CoA it wants to use for this flow.

    The main drawbacks of this approach afaict is that the first
    packets will flow through the primary CoA, but i guess these
    should not be too many packets and that there is a need to send a
    BU per each flow that will be sent through a CoA other than the
    primary one. Maybe these are not too many, don't know.

    The nice thing is that this is fully extensible since we never
    have to encode bits refering to the upper layer headers in the BU.

    Anyway, i thought it may be good to bring this up for discussion.

    Regards, marcelo

    _______________________________________________
    MEXT mailing list
    MEXT at ietf.org <mailto:MEXT at ietf.org>
    https://www.ietf.org/mailman/listinfo/mext