IETF58 PWE3 Working Group StructureAgnostic TDM over Packet draftietfpwe3satop00'txt Summary and Ope

1
IETF-58PWE3 Working GroupStructure-Agnostic TDM
over Packetdraft-ietf-pwe3-satop-00.txtSummary
and Open Issues

• Sasha Vainshtein, Yaakov Stein

2
4 months ago in Vienna

• From the PWE3 Session Minutes
• Yaakov Stein and Sasha Vainshtein to produce a
  joint unstructured draft by 1st September
• Yaakov Stein and Sasha Vainshtein invited to
  re-present their structured work as individual
  drafts
• This decision resulted in

3
SAToP-00

• A common draft co-edited by Yaakov and Sasha
• A new term - "struture-agnostic" coined to
  reflect the essence
• Co-authors
• Motti Anavi, Tim Frost, Eduard Metz, Prayson
  Pate, Akiva Sadovski, Israel Sasson, Ronen
  Shashoua
• Posted as an individual submission on 04-Sep-03
• Multiple positive responses on the list
• Some issues raised during the discussion
• Adopted as a WG item 23-Sep-03
• Open issues listed in the re-submitted text
• WG input solicited for their resolution

4
SAToP-00 (2)

• Supports E1/T1/E3/T3 TDM bit-streams
• Fixed-size chunks of the TDM bit stream
  packetized
• Default packet size specified per service
• Any packet size can be used if agreed upon
• A control word is mandatory
• The generic PWE3 CW is used
• Sequence number follows the RTP rules
• Circular 16-bit space without any gaps
• Two flags
• L - forward indication of the local AC failure
• Payload MAY be omitted
• R - backward indication of PSN failure

5
SAToP-00 (3)

• RTP header can be optionally used
• Positioned differently for IP and MPLS PSN
• IP immediately PRECEDES the CW to resemble the
  classic RTP
• MPLS immediately FOLLOWS the CW to be ECMP-safe
• Same sequence number used in both the CW and RTP
  header
• Two modes of time-stamping
• Absolute - the time-stamping clock is the line
  clock of the local AC
• Differential - the time-stamping clock is a
  high-quality clock available at both ends of the
  PW

6
Issue "Octet-Aligned T1"

• Raised by Ron Cohen and Yaron Raz
• Modern NSPs often extract (unstructured!) T1
  mapped to 25 timeslots within a (structured) E1
• The mapping is defined in ITU-T G.802 (Annex B)
• Relevant NSPs SONET/SDH mappers etc.
• Integration with SAToP is problematic
• Carrying an entire E1 using SAToP is BW-expensive
• De-mapping T1 from E1 requires additional HW
• Proposals (WG input is solicited)
• Add a special mode in SAToP?
• Define a new "octet-aligned T1" service?
• What else?

7
Issue "T3 AIS"

• Raised by Alex Conta
• T3 AIS cannot be detected/generated in a
  structure-agnostic way
• Requires T3 framing
• No such problem with the rest of the services
  (E1, T1, E3)
• AIS is important for SAToP
• Detection allows BW saving
• Generation is desirable in case of AC and PSN
  failures
• The current solution ("all ones") indicates a
  problem but not the specific problem
• Proposal (WG input is solicited)
• Direct indication of TDM validity!

8
Issue "A reference to the EF PHB"

• Raised by the Editors
• Using EF PHB in SAToP PWs seems natural
• E.g., see draft-baker-basic-diffserv-classes
• Appropriate limitations can be explicitly defined
• The DiffServ WG co-chairs objected to naming any
  specific PHB in the QoS section of TDM drafts
• This authoritative ruling has been accepted
• The doubts remain
• The WG input and the ADs guidance are solicited
• The DiffServ WG has concluded

9
What Remains to Be Done

• Resolve the open issues
• Allocate the service type code points for the
  SAToP-supported services
• Provide a SAToP MIB
• Will be done in a dedicated draft
• Add RTP-specific parameters to the PWE3 control
  protocols
• Will be done as
• A single dedicated draft in this WG?
• Two control protocol-specific drafts in different
  WGs?
• Go to the WG Last Call before the next IETF

10
Questions?