haas@ece.cornell.edu wrote:

One example that comes to mind is HSLS (the main example used in your draft).Hi Richard,

Thanks for your comments on the draft.Maybe another term should be introduced, such as "optimally scalable", that

indicates a given method is more scalable than any other possible method

(when an absolutely scalable method does not exist).

The term "optimal scalability" sounds interesting, but I am not sure how

it could be practically applied. In other words, I am not sure how one can

state that there cannot exist another method that is more scalable. If you

have any example in mind, please bring it up - it would help to understand

this term.

Quoting from the abstract of the MobiHoc 2001 paper that presented HSLS:

"This algorithm is shown to have nearly the best possible asymptotic overhead for

any routing algorithm - proactive or reactive."

So if the word "nearly" could be removed, then HSLS would be "optimally scalable"

in the number of nodes, with respect to the total overhead metric and the network

model defined in the paper.

In general, if one could *prove* that method A is "more scalable" than any other method (for

a given metric and network model), then one can say that method A is "optimally scalable".

But this is really a minor point, since your draft already defines "more scalable", so it almost

goes without saying that a method that is "more scalable" than any other method is

"optimally scalable". One reason I like the term "optimally scalable" is because, in some cases,

an "absolutely scalable" method may not exist, in which case the ultimate goal is to find an

"optimally scalable" method.

Regards,

Richard

In essence, what I am saying is that if a method A scales, say, with the

number of nodes as O(N^a), and a<0 (i.e., A is not absoultely scalable),

then, in principle, there always can be another method B which scales better

with N; i.e., O(N^b), where a<b<0.

Does this make sense?

Zygmunt.

=-=-==-=-=-=

~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-

Prof. Zygmunt J. Haas http://people.ece.cornell.edu/haas

Wireless Networks Laboratory http://wnl.ece.cornell.edu

School of Electrical Engineering

Cornell University tel: +1-607-255-3454

323 Frank Rhodes Hall fax: +1-607-255-9072

Ithaca, NY 14853 e-mail: haas@ece.cor nell.edu

U.S.A

WNL web page: http://wnl.ece.cornell.edu

~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-

On Wed, 17 Sep 2003, Richard Ogier wrote:Hello,

I can't attend the ANS meeting tomorrow, but I have a few comments on the

scalability draft <draft-irtf-ans-scalability-definition-00.txt>.

I like the definitions of absolute and relative scalability, and the

fact that all metrics

must be considered for absolute scalability (but a metric must be

specified for

relative scalability).

I think there are many important environments for which an absolutely

scalable

method does not exist, e.g., if the number of hops between a source and a

destination is not bounded by a constant. In these environments, relative

scalability must be considered, and the goal will be to show that a

given method

is more scalable than other methods.

Maybe another term should be introduced, such as "optimally scalable", that

indicates a given method is more scalable than any other possible method

(when an absolutely scalable method does not exist). < br>

Richard

_______________________________________________

Ans-research mailing list

Ans-research@ietf.org

https://www1.ietf.org/mailman/listinfo/ans-research

_______________________________________________

Ans-research mailing list

Ans-research@ietf.org

https://www1.ietf.org/mailman/listinfo/ans-research

_______________________________________________ Ans-research mailing list Ans-research@ietf.org https://www1.ietf.org/mailman/listinfo/ans-research From exim@www1.ietf.org Tue Sep 23 23:01:37 2003 Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged)) by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA21107 for