RE: [802.3ae] Wan Interface Sublayer
Gary
There has been a lot of "talk" about "jitter specifications". In a
conversation with some of our technology people that have been around a
long time, I fond out something interesting about why the SONET and SDH
specifications are what they are. The SONET and SDH specifications are
based around the requirements for electrical multiplexing of not only
digital signals but "analog" signals as well. The old T0/T1/T3 signals
were analog signals that were "bit stuffed" to provide alignment pointers
within the SONET transmission system. The jitter specifications were real
tight because of the compounding of the "adjustments" cause reliability
issues with the "analog" client signals.
When going through amplifiers the issues of jitter become compounded. When
going through SREs the clock recovery and regeneration cleans up any
"jitter". In a service deployment, a service provider that does not at
least have an SRE as a service demarcation is not going to be very serious
about their business. In looking at the way DWDM systems will be deployed
in order to provide service demarcation, I doubt that "jitter" will be a
real issue. The upper "end" of that service demarcation will be an LRE (in
the politically correct terminology of 802.3, ELTE) to demark the
customer's line overhead as well as their section overhead. This is not
much different than the way it is done between service providers, except
that the service providers generally use full sub-rate multiplexing LTEs to
do transmission service handoff.
Thank you,
Roy Bynum
WorldCom
At 05:17 PM 8/27/01 -0400, Gary Nicholl wrote:
>Roy,
>
>One difference might be that OC-192c POS interfaces are fully compliant to
>SONET/SDH jitter specifications (in terms of generation, transfer and
>tolerance). The two parameters that are important with regard to interop
>with long haul DWDM systems are obviously jitter generation and jitter
>tolerance. If I understand correctly the 10GbE WAN PHY is not currently
>aligned with SONET/SDH jitter specs. Any thoughts about the impact of this
>on the interop with long haul systems ? I guess it was always my
>understanding that one of the functions of this mysterious ELTE was to
>convert from WAN PHY jitter specs to SONET/SDH jitter specs?
>
>
>Gary Nicholl .............
>
>
>
>At 10:29 AM 8/27/2001 -0500, Roy Bynum wrote:
>
>>Mike,
>>
>>We are currently deploying OC192c POS over long haul systems using only
>>amplifiers and SREs. The network management of the POS systems at each
>>end of the SONET span is separate from the SONET span. The POS systems
>>are "free-running" relative to SONET synchronization. Fault protection
>>is being handled by the SREs. I do not see any difference in deploying
>>10GbE WAN PHY in the same way.
>>
>>Thank you,
>>Roy Bynum
>>
>>At 04:56 PM 8/24/01 -0500, Ayers, Mike wrote:
>>
>>
>>> > From: Roy Bynum [mailto:rabynum@xxxxxxxxxxxxxx]
>>> > Sent: Friday, August 24, 2001 01:09 PM
>>>
>>> > Being a SONET transmission service provider with OC192 SREs
>>> > already in the
>>> > transmission network. With the WAN PHY at +/-20PPM, your
>>> > possibility #1 is
>>> > correct. The SONET SREs (read SONET class B regenerators)
>>> > can handle the
>>> > WAN PHY and provide section level performance monitoring and
>>> > protection.
>>>
>>>
>>>
>>> > From: James Colin [mailto:james_colin_i@xxxxxxxxx]
>>> > Sent: Friday, August 24, 2001 12:14 AM
>>>
>>> > Option (4) is the correct answer. SONET and WIS must
>>> > stay on their own network. A network device called
>>> > "ELTE" is doing the bridging between the WIS net and
>>> > the SONET.
>>>
>>>
>>> Hmmm - I think a little more discussion on this is in order, and by
>>>minds greater than mine. My interest is this: if it is possible to route
>>>WIS directly over a SONET network, then the management model for WIS becomes
>>>much more involved, since it must support management by either the SONET
>>>model or the traditional ethernet model (two very different models).
>>>However, if an ELTE is required betrween the two, it effectively separates
>>>the networks such that there is no overlap nor need for two management
>>>models; ethernet being used on one side of the ELTE and SONET on the other.
>>>It does not matter, from my perspective, whether an ELTE gets used - what
>>>matters is whether the ELTE is required. Is it possible that the SRE is
>>>acting as an ELTE in Roy's scenario?
>>>
>>>
>>> Thanks,
>>>
>>>/|/|ike
>