Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Why an optical PHY makes 10GbE a non-LAN protocol




Roy,

I think we're getting too far off on a tangent, but briefly:

IBM 3044 channel extenders provide no modifications to channel operations support,
performance monitoring, management, etc. since the attachment to the host processor is via
the original parallel channel. Response timings were modified in certain areas of the
protocol to ensure compatibility only with parallel channel timings. However, the response
timings provided no management benefits.

IBM's ESCON(tm) channels (native, not channel extension) on the other hand, provided
significant extensions to operations support, performance monitoring, management, etc. since
they were now natively attached to the host processor. In addition switches and control
units to which peripherals were attached were accessible through in-band management
protocols. Fibre Channel, heavily based on ESCON, and separately Ethernet followed similar
paths.

My assertion is that Ethernet has sufficient management hooks in the existing protocol and
world class interoperable management support at all OSI layers to support any environment in
which it may venture. This includes MAN and WAN. Furthermore, by addressing these new
management requirements in a standard interoperable manner the resulting solution may end up
being both compatible with the existing infrastructure and more cost effective in its native
form.

Best Regards,
Rich

--

Roy Bynum wrote:

> Rich,
>
> The processing by the ESCON channel extenders provided modifications to the responce
> timing, operations support, performance monitoring, management, etc. that was not in the
> local only, channel attach link, directly from the mainframe.   By understanding the
> need for "modifications" to the channel attach link process, IBM was able to come out
> with operationally dependable distance extensions.  All I am trying to do is get others
> to be aware of the need to look at the issue of distance extension from a
> non-traditional aspect, the same way that IBM did.  They made some modifications to the
> way that the channel attach processes worked at what would be equivilent to the L1 link
> level.  I am asking the 802.3 HSSG to be open to the same type of thinking.
>
> Thank you,
> Roy Bynum,
> MCI WorldCom
>
> Rich Taborek wrote:
>
> > Roy,
> >
> > I know that you have made several arguments to the point that Gigabit Ethernet is
> > not a LAN because of its 1000BASE-LX variant which supports extended distances. I'd
> > like to respond with some relevant history which I am personally familiar with:
> >
> > Way back in 1988 or so (+ or - one year) I was working at IBM designing mainframe
> > channels. One of the neatest products I've had the opportunity to work with at the
> > time was a 3044 Channel Extender. This product consisted of two physical boxes, one
> > connected to the Bus and Tag cables of a mainframe channel interface, the other
> > connected to the Bus and Tag cables of the first in a series of peripheral devices,
> > other devices being daisy chained to the first device. A fiber optic cable connected
> > the two 2 3044 units. In this fashion the existing 120 meter Bus and Tag cable limit
> > could be extended to 2 km over 62.5 um MMF and 3 km over 50 um MMF. The line rate
> > was 200 Mbps, the encoding was 8B/10B, and the transmitter was an LED. The 3044 was
> > an excellent technology evaluator for IBM's current fiber optic-based mainframe
> > channels dubbed ESCON (TM). ESCON channels, first introduced in 1990 and the current
> > workhorse, use the same basic PHY but different framing than the 3044 channel
> > extended. One other significant addition is that ESCON channels also supported a
> > laser-based PHY variant, extending the supported ESCON distance to 20 km with a 1300
> > nm laser source. Up to three links could be arranged in line through two switches,
> > forming a 60 km link between mainframe channel to peripheral.
> >
> > Other LAN protocols like FDDI supported similar extended LAN distances. Fibre
> > Channel broke the 1 Gbps barrier with a completed standard in 1994, supporting
> > distances of up to 10 km in ANSI X3.230-1994 (FC-PH). The GbE 1000BASE-X PHY is
> > based upon the FC-PH PHY.
> >
> > The point in all the prior rambling is the GbE was not the first to venture into the
> > extended LAN (call it MAN or WAN, whatever) space. Other proprietary or standard
> > solutions may have gotten there earlier than even the 3044. I'm sure that the IEEE
> > 802.3 has no shortage of communications systems historians.
> >
> > Existing LAN and SAN data communications protocols including Ethernet, FC, HIPPI,
> > FDDI have done more than extremely well at meeting the needs of the environments for
> > which they were designed from a cost, functionality, reliability, manageability,
> > operations, serviceability, configuration flexibility, salability (I could go on and
> > on) point of view. The WAN marketplace, and perhaps the MAN marketplace as well, is
> > new and unfamiliar territory to many from the LAN and SAN community.
> >
> > I don't find it altogether unreasonable, that given the proper requirements,
> > guidelines, and historical perspective, that 10 Gigabit Ethernet can assemble a
> > standard that addresses all LAN and extended LAN environments as did its
> > predecessors as well as venturing into the MAN and WAN space while simultaneously
> > providing compatibility with all of the existing MAN and WAN infrastructure.
> >
> > What's wrong with the above picture?
> >
> > --
> >
> > Roy Bynum wrote:
> >
> > > ... (text deleted)
> >
> > > A WAN protocol has the ablity to work over extended distances;
> > > a LAN protocol does not.  Does full duplex 802.3 have the ability to work over
> > > extended distances?
> > >
> > > This is a door that has already been opened.  802.3 was altered into an extended
> > > distance, WAN, protocol by making it full duplex.  It may not have been the
> > > intention of the 802.3 WG to do so, but that is what was accomplished.  Adding
> > > an extended distance PHY sanctioned 802.3 for use over extended distances,
> > > either over privately owned facilities, or over commercial services facilities.
> > >
> > > I simply want to point out that what was accomplished by the 802.3 WG was
> > > greater than what they intended.  They opened the extended distance, WAN, market
> > > place to Native data, 802.3.  This is going to have more of a major impact on
> > > long haul data communications that it will have on the LAN environment, which
> > > already has Native data, 802.3.  The over all long term result will be lower
> > > cost data communications over extended distances, WANs as well as LANs.
> > >
> > > Thank you 802.3 WG,
> > > Roy Bynum,
> > > MCI WorldCom
> > >
> > > Drew Perkins wrote:
> > >
> > > > Roy,
> > > >         I believe that the difference between LAN and WAN (including MAN)
> > > > protocols has been mentioned in another very recent note. The key difference
> > > > is whether or not it is the same administration that owns and operates both
> > > > ends of the communication as well as the media in between. This is typically
> > > > the case with LANs. WANs, on the other hand, are typically owned, operated,
> > > > and managed by a service provider. Because the service provider needs to
> > > > have good tools to know when they are not providing good service, and to
> > > > locate the problem causing bad service, and to restore service while there
> > > > is a problem, WAN protocols tend to have facilities to provide superior
> > > > OAM&P compared with LAN protocols. Flavors of Ethernet do not currently have
> > > > the mechanisms for providing these tools.
> > > >
> > > > There may also be significant differences between MAN and WAN protocols. In
> > > > both the SONET/SDH worlds and DWDM worlds, the protocols used over MANs and
> > > > WANs may differ significantly. MAN protocols are often built to favor
> > > > simplicity and performance over efficiency. Hence the use of UPSRs. WAN
> > > > protocols are more often built to favor efficiency over simplicity and
> > > > performance due to the significantly higher costs. Hence the use of 4 Fiber
> > > > BLSR.
> > > >
> > > > Drew
> > > > ---------------------------------------------------------
> > > > Ciena Corporation                 Email: ddp@xxxxxxxxxxxx
> > > > Core Switching Division                 Tel: 408-865-6202
> > > > 10201 Bubb Road                         Fax: 408-865-6291
> > > > Cupertino, CA 95014              Cell/Pager: 408-829-8298
> > > >
> > > > -----Original Message-----
> > > > From: owner-stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> > > > [mailto:owner-stds-802-3-hssg@xxxxxxxxxxxxxxxxxx]On Behalf Of Roy Bynum
> > > > Sent: Sunday, June 27, 1999 6:15 AM
> > > > To: Peter_Wang@xxxxxxxx
> > > > Cc: stds-802-3-hssg@xxxxxxxx
> > > > Subject: Why an optical PHY makes 10GbE a non-LAN protocol
> > > >
> > > > Peter,
> > > >
> > > > I have been trying to figure out what the distinction between a LAN protocol
> > > > and
> > > > others.  There was the "assumption" made with the development of 1000BaseLX
> > > > that the
> > > > moderate power and 13xx wavelength specification would made it a MAN
> > > > protocol. It
> > > > did not, it only opened the door for it to no longer be a LAN only protocol.
> > > > In
> > > > optical networking with optical PHYs, there is no such thing as a distiction
> > > > between
> > > > a MAN and a WAN protocol, only between LAN and non-LAN protocols.
> > > >
> > > > In re-thinking the distiction between the protocols that are restricted to
> > > > short
> > > > distances, and those that are not, I may have found the answer.  Direct
> > > > responce
> > > > protocols, such as ESCON have a specific acknowledge responce time on each
> > > > data
> > > > block.  The transaction concurrency responce time of ESCON resticts it to a
> > > > limited
> > > > distance.  The IBM standard has some specific distance limitations to it.
> > > > That
> > > > specific distance limitation, based in data block acknowledge responce time,
> > > > limits
> > > > ESCON to being a LAN standard.  Changing the PHY by making it optical with
> > > > high
> > > > launch power will not make into a WAN protocol, because the limitation is
> > > > not in the
> > > > PHY, it is in the protocol itself.
> > > >
> > > > Because of the latencies involved with the slow responce time of end systems
> > > > and
> > > > mulitiple segments linked with repeaters, the original 802.3 did not have
> > > > restrictive frame acknowledge responce times.  At the time, 802.3 was
> > > > distance
> > > > limited by the distances of coax cable. When full duplex 10BaseT came into
> > > > being,
> > > > 802.3 became a point to point protocol, not much different from any other
> > > > point to
> > > > point protocol.  Some vendors even developed optical conversion and bridge
> > > > interfaces that were optical.
> > > >
> > > > When 802.3 was increased to full duplex 100 mb speeds and given an optical
> > > > PHY, the
> > > > distance limitations were offically removed.  Full duplex 100BaseSX is a
> > > > non-LAN,
> > > > point to point protocol.  Some people even implemented 100BaseSX as in a MAN
> > > > configuration using optical wavelength converters.  Only economics and
> > > > access to
> > > > long distance fiber prevented 100BaseSX from becoming a WAN protocol at the
> > > > very
> > > > first.  Full Duplex 1000BaseLX offically gave sanction to non-LAN 802.3.
> > > > With
> > > > optical wavelength converters, even 1000BaseSX does not have any LAN
> > > > distance
> > > > limitations.  Only the "assumption" that 802.3 was a LAN protocol did not
> > > > allow the
> > > > 802.3 WG to reconize that they were creating a non-LAN standard.
> > > >
> > > > The lack of restrictive frame acknowledge responce timers will also make
> > > > serial
> > > > optical 10GbE a non-LAN protocol by default.  Optical amplifiers will allow
> > > > implementations of 10GbE that will go at least 600km.  In the optical
> > > > networking
> > > > environment there is no difference between a MAN or WAN at the protocol
> > > > level.
> > > > Because of physical installation limitations, multi-fiber, parallel, MMF,
> > > > 10GbE may
> > > > be a LAN only implementation.  In spite of the "assumption" that 802.3 is a
> > > > LAN
> > > > protocol, serial optical 10GbE will not be.  Serial optical 10GbE will be a
> > > > MAN/WAN
> > > > protcol.
> > > >
> > > > Thank you,
> > > > Roy Bynum,
> > > > Optical and Data Network Technology Development
> > > > MCI WorldCom
> > > >
> > > > >
> >
> > --
> >
> > Best Regards,
> > Rich

-------------------------------------------------------------
Richard Taborek Sr.    Tel: 650 210 8800 x101 or 408 370 9233
Principal Architect         Fax: 650 940 1898 or 408 374 3645
Transcendata, Inc.           Email: rtaborek@xxxxxxxxxxxxxxxx
1029 Corporation Way              http://www.transcendata.com
Palo Alto, CA 94303-4305    Alt email: rtaborek@xxxxxxxxxxxxx