Re: [EFM] Re: OAM - To side-band or not to side-band
Matt,
When we were working on 10GbE, the issue of definition of "deterministic"
came up. I could not find it in any dictionary, including an unabridged
Webster. I did find a definition of "deterministic at
http://hissa.nist.gov/dads/HTML/deterministc.html. I also found a
definition of "non-deterministic" at
http://hissa.nist.gov/dads/HTML/nondeterministic.html. The difference
between the two definitions that can be related to this discussion is
whether there is single predetermined or multiple non-predetermined delays
that can be returned. From your comment about a fixing a minimum and
maximum , which indicates that at any one instant there are multiple
non-predetermine delays that can be returned, is actually non-deterministic.
Thank you,
Roy Bynum
At 01:12 PM 1/27/2002 -0500, Matt Squire wrote:
>According to Webster, deterministic comes from determine which means 'to
>fix the boundaries.' Hence, setting a minimum and maximum is quite
>deterministic.
>
>Nothing we could do would operate on 'any specific granular time frame',
>for a time frame can always be chosen which is less than 1-bit time. I
>know you have time frames you find desirable - other folk have other
>views on desirable time frames. Time frame granularity is certinly one
>of the differentiators between the OAM transport proposals and should be
>considered by folks when making their evaulations of the transport
>proposals.
>
>I'm completely unclear on what we approach 802.3ah OAM will not be
>taking.
>
>Roy Bynum wrote:
> >
> > Matt,
> >
> > The moment that you say minimum and maximums, you have said that it is not
> > deterministic from the service provider viewpoint for any kind of service
> > than "Internet" related. Deterministic means that at any specific
> > granular time frame, a specific level of performance monitoring is
> > available. On today's infrastructure, where it is managed at all, that
> > time frame is measured in microseconds. Where infrastructure is not
> > managed, the facility is managed through the services using SNMP embedded
> > in the customer revenue stream. Sometimes ATM is used to create an
> > dedicated channel at L2 for the SNMP, but it is still embedded as part of
> > the service revenue stream. I think that we have agreed that the 802.3ah
> > OAM will not be taking that approach.
> >
> > Thank you,
> > Roy Bynum
> >
> > At 12:09 PM 1/27/2002 -0500, Matt Squire wrote:
> >
> > >All of the OAM transport proposals are completely contained within 802.3
> > >and do not rely on implementation specifics. Also, all of the transport
> > >proposals have easily derivable deterministic minimum and maximum
> > >performance (bps).
> > >
> > >Bob Barrett wrote:
> > > >
> > > > Gentlemen
> > > >
> > > > I would assert that anything we define for OAM transport should also be
> > > self
> > > > contained within 802.3 and not rely on implementation specifics in
> order to
> > > > work deterministically.
> > > >
> > > > Non-deterministic management transport will not meet the broad market
> > > > acceptance criterion imho.
> > > >
> > > > Bob
> > > >