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

RE: [MIP-QOS] RE: MIP-QOS MIP QoS Mailing List is Active Now



Title: RE: MIP-QOS MIP QoS Mailing List is Active Now
Gwendal,
-----Original Message-----
From: ext Gwendal Le Grand [mailto:Gwendal.Le-Grand@lip6.fr]
Sent: Thursday, April 19, 2001 5:15 AM
To: mip-qos
Subject: RE: [MIP-QOS] RE: MIP-QOS MIP QoS Mailing List is Active Now

Glenn,
 
I agree with what you said here, i will just make one comment about using FEC and unequal protection.  believe this is definately a good idea. But maybe it should not be required for all "important" paquets. I think this implies some channel estimation techniques to be used (isn't it out of the scope of this work ?)
[Morrow, Glenn ] 
It seems to me that some signaling has to be able to communicate to the access networks that this function is to be done. The signaling to do this can be triggered via interactions with Layer 5 (SIP, H.323, etc..) which would imply some special access specific and security association with the access routing equipment or the parameters can be included as part of some end to end reservation signaling. Including the media type in the end to end reservation signaling allows a pure host to host communication with QoS and without the need for a proxy at all or at least not associated with the visited access network.If an access link layer does not support the function, it simply ignores the parameter. So with these points in mind, I think it should be in the scope of NSIS.

Moreover, we have" to be careful about not having the same problem as with IEEE802.11 where all the signaling is carried at a minimal rate for a better protection. In that specific case, that degrades performance when higher rates for the goodput are to be used. My point here is that FEC is probably more flexible and achieves better performance but it should be only used when the channel is bad.
[Morrow, Glenn ] 
I believe that some more work will eventually be done to the 802.11 MAC to provide CoS similar to the ethernet work. I believe an SBM will work for the bandwidth problem. I suspect that doing unequal protection on 802.11 would be overkill. If you examine the business aspects from a monetary perspective of 802.11 and cellular, you might begin to see that including a media type in the reservation signaling is more justified than not.
Another point about wireless channels (here i am thinking especially about wireless LANs) : some people say that the channel is often good (and no protcetion is required) and sometimes very bad (and no data can be transmitted). What do you think about this statement (to what extent is it true or false) ? Then, if it is really true, FEC will not improve performance.
[Morrow, Glenn ] 
If the cost of wireless LAN spectrum was significant and the primary  traffic type was expected to be voice, I would consider it. Neither is true, though, so I wouldn't ever consider it. I do think some CoS work could be done to 802.11 analogous to the .p work on ethernet which could significantly improve it. I wouldn't use unequal protection schemes though.
 
Thanks
 
Gwendal