RE: Example of LOM usage

[Dimitry Haskin <dhaskin@axiowave.com>]

Francois,

Correct, these are points that I'm making.  A minor correction is that I'm
not claiming that "most SPs won't need per-CT-per-link ratios". Nevertheless
the second bullet in point 2 is largely correct.

There seems to be a persistent impression that use of LOM (as currently
specified) is absolutely necessary when different per-CT-per-link
overbooking ratios are required whereas the quite opposite is true in
practice. This level of confusion that LOM creates seems disproportional to
its dubious value. This is why removing it from the base spec appears to be
a desirable way to proceed.

Dimitry

> -----Original Message-----
> From: Francois Le Faucheur (flefauch) [mailto:flefauch@cisco.com]
> Sent: Thursday, June 05, 2003 9:15 AM
> To: Dimitry Haskin; Lai, Wai S (Waisum), ALABS; te-wg@ops.ietf.org
> Cc: Ash, Gerald R (Jerry), ALABS
> Subject: RE: Example of LOM usage
> 
> 
> Dimitry,
> 
> >> I understand that a provider may want different per CT 
> >> overbooking ratios on
> >> different links. However I still fail to understand why 
> LOM would be
> >> required to provide a good service in such a configuration. 
> >> Can you clarify?
> >> 
> 
> I believe we have already discuss when LOM is required or not and
> concluded on that. Quoting from your message:
> >> Francois,
> >> 
> >> Thanks for your clear explanation. What has escaped me 
> >> before is that LOM is
> >> useful only for those cases where both of the following 
> must be true:
> >> 
> >>  1) CTs of the same link have different overbooking ratios.
> >>  2) The amount of reservations in one CT must precisely 
> >> effect reservable
> >> bandwidths in other CTs of the same link.
> >> 
> >> If I am correct, then it is quite possible to support 
> >> different overbooking
> >> ratios per CT per link without LOM, provided that 2) is not 
> >> required (or at
> >> least not with the same level of precision that LOM can provide).
> >> 
> >> Thanks,
> >>  Dimitry
> 
> To put the same thing in a different way:
> 	- if you don't need per-CT-per-link ratios, you do NOT need LOM
> 	- if you need per-CT-per-Link ratios and can tolerate some
> inaccuracy in how a CT affects the global constraints (which 
> I think you
> also refer to as allowing some degradation on some CTs in your
> discussion with Sanjaya), then you do NOT need LOM.
> 	- if you need per-CT-per-Link ratios and want accuracy in how a
> CT affects the global constraints, then you do need LOM.
> 
> So the points I think you're getting at are that:
> 	1) needing per-CT-per-Link ratios does not always mean you need
> LOM
> 	2) from your experience:
> 	- most SPs won't need per-CT-per-link ratios
> 	- the few SPs who may need per-CT-per-link ratios are likely to
> tolerate some inaccuracy in bookeeping over the aggregate constraint,
> Therefore the case for LOM is very slim.
> 
> Right?
> 
> Cheers
> 
> Francois
>