RE: Fw: [RPRWG] CRC check in each node?
Hello Reuven
This is an interesting issue.
We could leave the span management (i.e., to identification of possible
faults/degradations in the segment) for the PHY layer, and RPR could only
perform the Header CRC check so that we don't misroute packets. However, if
we don't count payload errored packets we don't know what is the % of the
traffic that is error-free. I think the service provider will want to
monitor SLA, and will want to know how many packets it is delivering and
what is the % of the errored packets being delivered. Although we don't
need to check for payload CRC at every node, it could be performed only
once, at the edge of the RPR network, before delivering it to the client.
Now I have another question to the group: if we leave span management for
the PHY layer to do (what I think is the best thing to do), how do we check
the communication between PHY and RPR inside the box? If I have constant
traffic being passed to RPR, than it is fine because I see that the packets
are being delivered. However, how do I know if I am not receiving any
traffic from my PHY because I may have a problem between the PHY and RPR
(the PHY is not going to say anything to RPR) or simply because I don't
have any traffic? ATM has a OAM flow being sent in parallel to solve
this...
Any thoughts on that?
Angela
Reuven Zeitak
<reuven.zeitak@nativenet To: "ieee 802.17 list" <stds-802-17@xxxxxxxx>
works.com> cc: (bcc: Angela T. Faber/Telcordia)
Subject: RE: Fw: [RPRWG] CRC check in each node?
07/03/01 01:46 AM
Hi Everybody,
I have been following this interesting thread. I want
to rock the boat a bit here to make sure all points have
been considered:
Question: Is a CRC on data really what we want?
Consider:
ATM only has a HEC and does not protect data,
it is left to the application.
In some sense RPR is just an encapsulation of user data
and has to be agnostic to the data content. As it was pointed
out here, Users may want to keep on getting the data even if
there are link errors. Maybe the application has a sophisticated error
recovery method. Maybe it has none. Is it the RPR layer's job to
provide the application an error indication?
We clearly need some kind of BER measure (in the SONET jargon)
to verify if the link is alive. Isnt the HEC good enough for that?
HEC errors are to be discarded at the next node anyway, so that
there is no issue of double counting an error.
reuven
----------------------------------------------------------------------------
-----------------------------
Reuven Zeitak PhD
Modeling And Algorithms
Native Networks
2 Granit St., P.O.Box 7165
Petah Tikva, Israel
Tel: +972-3-920-2800 x 875
Fax: +972-3-9210080
reuven.zeitak@xxxxxxxxxxxxxxxxxx
www.nativenetworks.com
The Native Way = Ethernet Simplicity + SONET Reliability