Re: Ethernet over WANs
Bill,
I understand the problems that you are having with the "dark recesses"
of
WAN/MAN transport services and systems. I understand the problems that
you are
having with GbE and DWDM. Over the last 10 years, I have experienced
many of
the same frustrations that you are expressing. It is extremely
challenging to
build data architectures that directly map over and directly use the
transport
systems and services. This is the reason that I have been bringing up
the issue
of operational support for 10GbE, first raising it at the meeting in
June. It
is my hope that by applying some wisdom to how the 10GbE protocol is
structured,
some of these issues will be addressed. I am looking forward to
continued
discussions with you on this topic.
Thank you,
Roy Bynum
MCI WorldCom
"Bill St. Arnaud" wrote:
> >
> >
> > Bill,
> >
> > I appreciate your comments, yet feel that little is served by merely
> > repetitively voicing a characteristic common to both the data
> > communications
> > industry and the telephony industry as though it were news.
>
> I concur and I have resisted making any specific postings to HSSG in this
> area unless specifically asked to make a comment by someone reading the
> archives as happened in this latest exchange of e-mails. I mistakenly
> copied HSSG in my reply which started this current dialogue.
>
> > Your previous
> > postings infer that the same patterns of vendor-specific network
> > management
> > platforms and systems that exist in the telephony industry are
> > being recommended
> > for 10GbE.
>
> I hope I have not given that impression. What I am trying to say is that
> OAM is a big challenge in terms of developing interoperable standards as we
> have seen in the SONET world and yet SONET networks work fine and there are
> many good management systems. For native GbE do we really want to get into
> this area of standards development??
>
> Thus, let me once again say that I am not proposing,
> > nor have never
> > considered recommending the same type or level of OAM processing
> > for 10GbE as
> > that used by the telephony industry.
>
> I have always understood that.
>
> > I am also not recommending
> > 10GbE adhere to
> > a particular router/data switch vendor's all encompassing network
> > management
> > platform. As technologists, we need to be careful to not confuse the OAM
> > functionality of a given protocol with particular vendor's
> > proprietary network
> > management system offering and market dominance strategy. Protocol
> > interoperability should not be confused with vendor proprietary product
> > strategies, regardless of how much we may be lobbied to believe otherwise.
>
> Agree 100%
>
> >
> > As for OAM interoperability, data and telecom history mutually
> > demonstrates that
> > when one stays exclusively with a particular vendor's management
> > system, most
> > likely it will not work optimally or -- in many unfortunate
> > experiences -- even
> > effectively with another vendor's system. In SNMP management systems, for
> > example, the commonly acknowledged alternative is to go to an
> > independent, third
> > party management system, such as "Open View" (trademark HP)
> > (While we are not
> > supposed to mention vendor's names here, this particular example
> > cannot be drawn
> > without it, as you're probably aware). Likewise, in the
> > telephony industry,
> > there are third party telephony management systems that similarly provide
> > effective cross-vendor OAM management. While avoiding the
> > mention of a vendor
> > name, there is such a vendor in Canada whose products are
> > successfully deployed
> > by carriers across multiple vendor systems.
>
> Without mentioning names that vendor does have OAM products that will manage
> across "some" other vendor systems, but not all. If you have some time I
> wish you could come and visit us and see the practical problems we face in
> terms of non-interoperability on both the SONET transport and OAM systems we
> have on IP/DWDM network. We do use systems like Open View for SNMP
> management and for giving us "views" into other proprietary systems. But
> even though our routers our DWDM coupled into the transport, for all
> practicality they have to be treated as totally separate network. All the
> normal funcationality of a proprietary or non-proprietary OAM is lost.
>
> The important issue is not about SNMP polling or management views but fault
> isolation and network performance. It is very fustrating to locate faults
> on a network when 2 different management systems have to be used on the
> network and both are giving conflicting information on terms of throughput,
> performance and where the fault lies.
>
> >
> > Much of the interoperability problem engendered by some standards
> > organizations
> > arises from vendor/governmental influence.
>
> Agreed
>
> 802.3 HSSG's clear
> > opportunity and
> > mission is to operate above these non-relevant distractions. A
> > useful example
> > of this unproductive influence is the diversity of wireless
> > standards is Europe.
> > Because international telephony standards are more regionally
> > drawn than those
> > of data, each vendor that has a major influence in a particular
> > nation can - to
> > a far greater degree than in other global technology sectors -- make its
> > recommendations a national issue with the international standards
> > organizations. Fortunately, while this multiplies the complexity of such
> > deployments, it does not mean that vendors can not
> > interoperate... much like the
> > recalled conflict between HDLC and SDLC. After all, if telephony
> > vendors could
> > not interoperate, we would not be able to have a phone
> > conversation between
> > North America and Europe, as a result of the difference between
> > uLaw and aLaw
> > voice encoding.
>
> Or SDH and SONET. So unfortunately carries have to deploy transcoders
>
> The IEEE 802.3 HSSG should appropriately leverage the most
> > effective and most interoperable of existing and new solutions, and to
> > continually broaden and evolve their deployability and utility,
> > rather than
> > building narrow, industry sector-segmented, bounded solutions.
>
> I agree. But as I said before my gut reaction is to aim for the minimalist
> common standard and the wait to see how the market evolves before setting
> standards on more complex interoperability issues.
>
> For example, there has been discussion on the list about the OIF
> recommending a digital wrapper to address the problem of bad splices and
> other physical layer issues. That is one possible solution, but not
> necessarily the only one. With the growing trend in customer owned fiber
> with a different set of operating requirements I would hate to be locked in
> a pre-defined solution that was intended for another marketplace and another
> time.
>
> >
> > I'll welcome our continued conversation in person at the September interim
> > meeting.
>
> I don't know if I will be able to attend.
>
> But please keep up the e-mail. It always hae been a useful and informative
> dialogue
>
> >
> > Thank you,
> > Roy Bynum
> > MCI WorldCom
> >
> >
> >
> > "Bill St. Arnaud" wrote:
> >
> > > Roy:
> > >
> > > Despite the fact there may be commonly agreed upon bits in the
> > SONET header
> > > I have yet to see any SONET OAM or SONET transport systems that are
> > > interoperable. In terms of interoperability they are not even
> > close to the
> > > network management capability of SNMP systems with MIBs, etc.
> > Most of the
> > > SONET network management systems have been designed to work closely with
> > > OSS.
> > >
> > > All I am saying is that SONET world cannot produce an interoperable OAM
> > > systems does the HSSG want to venture into that quagmire
> > without some first
> > > hand operational experience? I believe that form should follow
> > function in
> > > terms of any standards setting exercise.
> > >
> > > The OAM requirements may also be vastly different for native
> > 10GbE networks
> > > operated by ISPs and CLECs than those operated by carriers. In
> > fact I very
> > > much doubt that IXCs will ever operate native long haul GbE.
> > >
> > > Did you ever contact Andrew Matoga? If not please contact Mike
> > O'Dell - he
> > > is very familiar with the GbE network that UUnet built
> > >
> > > Bill
> > >
> > > Bill St. Arnaud
> > > Senior Director Network Projects
> > > CANARIE
> > > bill.st.arnaud@xxxxxxxxxx
> > > +1 613 785-0426
> > >
> > > > -----Original Message-----
> > > > From: Roy Bynum [mailto:rabynum@xxxxxxxxxxx]
> > > > Sent: August 8, 1999 7:54 PM
> > > > To: Bill.St.Arnaud@xxxxxxxxxx
> > > > Cc: stds-802-3-hssg@xxxxxxxx
> > > > Subject: Re: Ethernet over WANs
> > > >
> > > >
> > > > Bill,
> > > >
> > > > What are you trying to do? What is the issue here? We are not
> > > > talking about
> > > > core implementing core transport systems here. I do not think
> > > > that IEEE wants
> > > > to get into standardizing the core telephony transport systems.
> > > > Yes, the way
> > > > that different vendors have implemented some of the SONET/SDH OAM
> > > > processing is
> > > > not standardized, just as the extended SNMP MIBs of each of the
> > > > data vendors
> > > > are not part of the minimum standards. This is the way that
> > > > vendors distinguish
> > > > themselves and leverage their marketing. The SONET/SDH vendors
> > > > have implemented
> > > > a minimum standard interface through which they can interoperate.
> > > > What is so
> > > > different about that? The important thing is that the SONET/SDH
> > > > protocol IS
> > > > standardized. The important thing is that the 802.3 HSSG can
> > > > leverage that
> > > > technology at OC192C rate. What kind of confusion are you trying
> > > > to induce
> > > > here?
> > > >
> > > > Thank you,
> > > > Roy Bynum
> > > > MCI WorldCom
> > > >
> > > > "Bill St. Arnaud" wrote:
> > > >
> > > > > Roy:
> > > > >
> > > > > This list has touched on the topic many times, so I don't
> > want to bore
> > > > > people by reiterating the issues. But very simply there are
> > > > few vendors who
> > > > > have any kind of interoperable OAM on the SONET "transport"
> > > > side. I stress
> > > > > the key word is "transport". On the other side of an ADM there is no
> > > > > problem. We have worked with most of the major vendors and
> > > > have yet to see
> > > > > anything close to interoperability.
> > > > >
> > > > > Bill
> > > > >
> > > > > Bill St. Arnaud
> > > > > Senior Director Network Projects
> > > > > CANARIE
> > > > > bill.st.arnaud@xxxxxxxxxx
> > > > > +1 613 785-0426
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: Roy Bynum [mailto:rabynum@xxxxxxxxxxx]
> > > > > > Sent: August 7, 1999 5:12 PM
> > > > > > To: Bill.St.Arnaud@xxxxxxxxxx
> > > > > > Cc: Rohit Mittal; stds-802-3-hssg@xxxxxxxx
> > > > > > Subject: Re: Ethernet over WANs
> > > > > >
> > > > > >
> > > > > > Bill,
> > > > > >
> > > > > > As part of the common carrier industry that uses SONET/SDH all of
> > > > > > the time, and
> > > > > > as a person that was involved with codifying the communications
> > > > > > stack for Q3, I
> > > > > > wonder where you get the idea that OAM has not been standardized
> > > > > > within the
> > > > > > SONET/SDH protocol? Have your ever heard of the H bytes, the B
> > > > > > bytes, the K
> > > > > > bytes, the Z bytes, etc. All of these are part of the OAM of the
> > > > > > SONET/SDH
> > > > > > protocol.
> > > > > >
> > > > > > For 10GbE, not all of the OAM bytes will be used within
> > the interface
> > > > > > processing, thus reducing the cost of processing compared to
> > > > > > current SONET/SDH
> > > > > > systems. As far as proprietary implementations of external
> > > > > > management systems,
> > > > > > have you ever heard of extended MIBs. There is little
> > > > difference in the
> > > > > > additional features offered by data switch vendors and the
> > > > > > additional features
> > > > > > offered by transmission vendors.
> > > > > >
> > > > > > Either you have been badly misinformed by some vendor
> > > > somewhere, or do not
> > > > > > understand the overall issues of network management
> > within the carrier
> > > > > > industry. There are carrier (SONET) network management software
> > > > > > solutions that
> > > > > > will span multiple vendors. A major one of those vendors is
> > > > in Calgary,
> > > > > > Canada. It can be compared to the carrier version of the SNMP
> > > > > > network managers
> > > > > > that are well known by the data networking industry.
> > > > > >
> > > > > > The specific of the minimum OAM features that are implemented has
> > > > > > not entered
> > > > > > the discuss yet. I personally do not think that it is time yet.
> > > > > > Some of the
> > > > > > OAM features that are in SONET/SDH do not exist in common
> > > > data networking
> > > > > > protocols, so many of the people that reading from this reflector
> > > > > > are unfamiliar
> > > > > > with them. To enter into that discussion now would very much
> > > > confuse the
> > > > > > issues.
> > > > > >
> > > > > > Thank you,
> > > > > > Roy Bynum
> > > > > > MCI WorldCom
> > > > > >
> > > > > > "Bill St. Arnaud" wrote:
> > > > > >
> > > > > > > Rohit:
> > > > > > >
> > > > > > > I think you will see 3 different approaches in the
> > marketplace for
> > > > > > > delivering 10GbE over the WAN:
> > > > > > >
> > > > > > > 1. 10GbE over SONET where a proprietary OAM is done at
> > SONET layer
> > > > > > > 2. 10Gbe over wavelengths where a proprietary OAM is done out
> > > > > > of band in the
> > > > > > > optical channel
> > > > > > > 3. Native 10Gbe which would require some sort of new
> > OAM protocol
> > > > > > >
> > > > > > > The question that has been asked as since no common
> > > > standard has been
> > > > > > > developed for OAM in SONET, is a "standard" OAM required for
> > > > > > native 10GbE?
> > > > > > >
> > > > > > > Bill
> > > > > > > >
> > > > > > > >
> > > > > > > > Hi all,
> > > > > > > >
> > > > > > > > I recently subscribed to this list and was going through
> > > > the archives
> > > > > > > >
> > > > > > > > Some people have mentioned about sending ethernet over
> > > > MAN/WAN links.
> > > > > > > > The question I have is Ethernet doesn't have OAM features
> > > > of SONET.
> > > > > > > > That is a critical feature for WANs (and not LANs)
> > > > > > > > How is that handled for 10GbE? Can you make intellegent
> > > > enough DWDM
> > > > > > > > elements to handle that. Doesn't that by itself add
> > > > overhead which is
> > > > > > > > comparable to SONET overhead. If so, then why is there so much
> > > > > > > > discussion
> > > > > > > > as to replace SONET by 10GbE.
> > > > > > >
> > > > > > > >
> > > > > > > > In packet over SONET, we use PPP which adds very little
> > > > > > > > overhead. Can't we use the same format for 10GbE over
> > > > SONET. In that
> > > > > > > > case we can define the 10GbE standard without worrying
> > > > about SONET .
> > > > > > > > Just my 2c.
> > > > > > > >
> > > > > > > > Thanks
> > > > > > > > Rohit Mittal
> > > > > > > > ~
> > > > > > > >
> > > > > >
> > > >
> >