call 7/7/2015 notes
Folks,
Quick notes below.
Start: 7/7/2015 6AM PDT
End: 7/7/2015 ~7:06AM PDT
Present:
Jouni Korhonen
Duncan Cockburn
Steinar Björnstad
Richard Tse
Richard Maiden
Sriram Devi
Xhafer Krasnigi
Kevin Bross
Peter Turnbull
Peter Ashwood-Smith
Mohammad Akhter
Actual Agenda:
* clarifications on RoE header flow_id
* RoE control packet (format)
* More on error handling
* AOB
Discussion:
* Jouni presents a contribution draft on RoE header & flow_id handling and semantics.
* Sriram asks whether flows are symmetric between REs and RECs.
* Jouni acknowledges that they should be, although in most cases it is viewed that
the other end (most likely the RE) is "responsible" for flow_id space. Jouni said to
to provide more clarifications on this later on.
* Jouni presents a contribution draft on RoE control packet format. Also the handling
of flow_id in the case of control packets were discussed (typically not used since
control packets are meant for all flows between SA-DA pair - however, this is at the
end dependent on the control packet subtype.
* Usage of the control packet subtype (up to 256 of those) presented - will be the first
byte of the payload.
* Xhafer asks whether presentations will be available. Jouni says he will post them to
the mailing list.
* Richard M. brings the error handling up again. The concern is that when Ethernet packet
gets dropped because of FCS mismatch that is different from what we have e.g. on
CPRI. Dropping Ethernet packet loses all IQ information whereas on CPRI the error
Is likely to go through undetected
* It was clarified again that even if cut-through is a defector standard that cannot be
He used because it is not a standard as such.
* Multiple checksums brought up: 1) for header and 2) for payload. Jouni points out
That this is unlikely to be acceptable within our PAR since it would assume new
Ethernet-level functionality.
* More work is needed to understand the implications to RoE when there is packet
drops on the link. Richard taking a ball on this.
* Error related discussion got still around how to protect against packet drops. Ring
topology and seamless redundancy brought up. Especially latter is too costly. If
there were alternative routes switch between connections is also likely to be
too slow.
* ORI intro will be postponed to a later call. Richard on this. Mohammad also points
out there is interesting things to look at what has been done (e.g. compression).
* Sriram asks whether compression is in scope.
* Jouni says that it is in a sense RoE has to be able to transport compressed IQ etc
but we are not going to develop/standardize the compression itself.
* Sriram points out that compression is going to affect how we define samples are
stored in the RoE packet. Jouni agrees.
* AOB: Jouni says the next call overlaps with IETF. Will be informed later whether
next call takes place or not.
* Jouni tells that more procedural information related to August virtual WG meeting
Logistics will be posted to the 1904 mailing list later this week.
Action Points:
New:
1) Richard Maiden to provide further information errors and behavior at Ethernet, and how that
maps/shows at radio level (maybe get more data from test equipment vendors).
OLD:
2) Kevin Bross to make a writeup of propagating error information/indications detected e.g. on the last CPRI link SFP module.
3) Jouni K to make a proposal for RoE Control packet e2e security baseline.
Deadline next telco (7/21/15)
- Jouni
--
Jouni Korhonen, CTO Office, Networking, Broadcom Corporation
O: +1-408-922-8135, M: +1-408-391-7160
--
Jouni Korhonen, CTO Office, Networking, Broadcom Corporation
O: +1-408-922-8135, M: +1-408-391-7160