RE: Reflecting new-MAM/SAM definition in diff-te drafts

Francois, Waisum and all

IMHO "max reservable bandwidth" should be used rather than
max-link-bandwidth to control the aggregate bandwidth in MAM.
Indeed max-link-bandwidth should be rather used to limit bandwidth of
one LSP.
Overbooking allows to take into account the fact that statiscally, not
all LSPs are simultaneously loaded, but we should also limit the
bandwidth of one LSP

So I agree with Francois regarding the use of max-reservalbe-bandwidth,
and I suggest to add a third rule that is 

An LSP of bandwidth B can be accepted if 
	B < max-link-bandwidth 

Note that this rule also applies to RDM

Regards 

JL

-----Message d'origine-----
De : Francois Le Faucheur (flefauch) [mailto:[email protected]]
Envoye : mardi 20 mai 2003 11:56
A : Ash, Gerald R (Jerry), ALABS; Lai, Wai S (Waisum), ALABS; Dimitry
Haskin
Cc : [email protected]
Objet : RE: Reflecting new-MAM/SAM definition in diff-te drafts


Jerry, Waisum and all,

>> > If yes, would you agree that the following rules would 
>> apply for the
>> > "new MAM" definition:
>> > 	o for each value of b in the range 0 <= b <= 7: 
>> > 		Reserved (CTb) <= BCb, 
>> 
>> This definition is OK, wherein the formula can be applied to 
>> either 'normalized' (by the overbooking factor) or 
>> 'unnormalized' bandwidth.
>> 

Yes, absolutely, in case the per-CT Local Overbooking Multipliers (LOM)
are used, then the formulas need to be applied to the normalised
bandwidth (I was only trying to get some agreement first on the
definition in the simpler case where per-CT overbooking is not used,
sorry I didn't make that explicit).

Anyway, this part of the formula would not change from current version
of MAM draft:

1)when per-CT LOMs are not used:
"o	for each value of b in the range 0 <= b <= (MaxCT - 1):
            		Reserved (CTb) <= BCb,"

1)when per-CT LOMs are  used:
"-	for each value of b in the range 0 <= b <= (MaxCT - 1):
            		Normalised(CTb) <= BCb,"

So I think we all agree with that part of the formula, right?

>> > 	o SUM (Reserved (CTc)) <= Max Reservable Bandwidth, 
>> > 		for all "c" in the range 0 <= c <= (MaxCT-1)
>> 
>> I agree with Wai Sum that this formula is incorrect for 
>> DS-TE, since the above formula only reflects the Max 
>> Reservable Bandwidth for the entire link, and does not 
>> reflect the per-CT overbooking factor.
>> 
>> Since there is a per-CT overbooking factor, this must be 
>> reflected in a summation of reserved bandwidth.  As such, 
>> use of the following formula would appear to more correct:
>> 
>>   SUM (Reserved (CTc)/overbooking(CTc)) <= Max Link Bandwidth
>>     for all "c" in the range 0 <= c <= (MaxCT-1)

Again, here I agree that the formula needs to be applied to Normalised
bandwidth when per-CT LOMs are used. 

Also, I think we are in agreement that we should include an explicit
condition in the MAM definition which constraints "SUM (Normalised
(CTc))" (ie to make explicit the formally "implicit" constraint). Am I
right that we agree on that?

So, the only divergence seems to be that when doing the check on SUM
(Normalised (CTc)):
	- you propose to use "Max Link Bandwidth"
	- I propose to use "Max Reservable Bandwidth"

Again this seems to be coming from different interpretations of these
two IGP parameters. 
Like Dimitry pointed out, to me the parameter which is meant to be used
for reservations is "Max Reservable Bandwidth". This what I read in the
IGP extension text. Also, as mentioned earlier, the TE implementations
that I am familiar with are currently all using "Max Reservable
Bandwidth" for aggregate admission control decisions and not "Max Link
Bandwdith".

Cheers

Francois


>> 
>> Thanks,
>> Jerry
>> 
>>