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

RE: Distance Objectives !!!!!


A good start. One question on item 3, though. If a "campus-to-campus", dark
fiber based MAN connection is supplied by an RBOC, shouldn't we count twice
the segment length as the length of a MAN link? In other words, shouldn't
we consider 8 kilometers a reasonable target for a dark fiber MAN link? The
placement of a repeater at a CO location may not be desirable or necessary
in dark fiber applications.

My (limited, subjective) observations with Storage Area Networks
applications (Fibre Channel link extenders) has been that 10 kilometers
covers about 90% of the cases.

And then there is the possibility that some dark fiber suppliers may not
use the RBOC infrastructure, and that their routes may be along electrical
lines, railway lines or pipelines.

Any thoughts?


Vipul Bhatt
Finisar Corporation
Phone:(650)691-4000 x113
Email: vipul.bhatt@xxxxxxxxxxx


> -----Original Message-----
> From: owner-stds-802-3-hssg-distance@xxxxxxxxxxxxxxxxxx
> [mailto:owner-stds-802-3-hssg-distance@xxxxxxxxxxxxxxxxxx]On Behalf Of
> Haile-Mariam, Atikem
> Sent: Friday, June 25, 1999 12:11 PM
> To: 'BRIAN_LEMOFF@xxxxxxxxxxxxxxxxxxxxxxxxxx'
> Cc: stds-802-3-hssg-distance@xxxxxxxx
> Subject: RE: Distance Objectives !!!!!
> Importance: High
> With some many reflectors, I haven't had a chance to catch up!!
> I haven't quite finished my research, but I propose a slight
> modification to
> Brian's proposal:
> 1) At least 300 meters on MMF for installed base
> 2) At least 500 meters on MMF for new grades of MMF  (this would very
> similar to what we developed for 802.3z in that we included additional
> higher cell bandwidth fibers)
> The logic for this has been discussed many times before, and so
> I don't want
> to "rehash" it here....(Also, Bruce's ad hoc will have a lot more to say
> about this in Montreal)....
> 3) At least 4 km's on SMF
> 4) At least 20 km's on SMF
> The logic for 3) and 4) is as follows:
> 3) A "typical" RBOC utilizes/deploys networks based on a "star" topology;
> with fiber running from the central office (CO) to a "drop" or building.
> AT&T, (and later Bellcore), conducted studies in 1964, 1974, and
> 1983 that
> measured min, mean, max distances from the CO to the termination
> point.  In
> the last study, (or at least I think it is the last study), the
> values were
> as follows:
> Min: 	186 ft
> Mean:	10,780 ft
> Max:	114,000 ft
> To each of these numbers you need to add a "typical" fiber run
> required to
> accommodate the distribution in the drop. That value is 1,888
> ft.  So mean +
> distribution drop =
> 10,780 ft + 1,888 ft = 12, 668 ft >>>> 3.85 km >>>> ~ 4 km
> Now, most people may be wondering why I would rely on such an old
> study...The reason I'm using this data is that the RBOCs have not
> consolidated CO's despite the ability to do so through the use of new
> digital switches. (Believe me, if they had consolidated, DSL
> would be a lot
> harder to implement than it already is...)
> 4) The "logic" for > 20 km's is a lot less "solid", but it goes something
> like this:
> CLECs - unlike RBOCs - deploy a "ring" topology. (A ring is arguably much
> more efficient since 1 CO can be used to service it). Now, depending on
> where you are in relation to the CO, the link length can "very short" or
> almost twice the length of the ring. (As an analogy, think of FDDI on how
> you would need to "loop back" in case of a failure)...
> Now for the numbers:
> The average CLEC route miles is 200 km's  Based on the average number of
> buildings - or buildings "passed" -  in the loop, we roughly have 50
> km/building....This of course is a mis-leading statistic since not every
> building is an even distance from the CO...The solution is to
> analyze some
> of the larger CLEC and the average building/termination runs.
> That figure is
> ~20 km's.
> I hope to compile a more formal presentation by next week that includes
> estimates of the amount of dark fibers in these networks.
> Regards
> Atikem Haile-Mariam
> P.S. Incidentally, I'm in full agreement with Ed in arguing that
> architectures should drive link length requirements - not components. (It
> just so happens, that based on what I saw in Idaho components
> may exist that
> fit "nicely" in these architectures).