Upload
aileen-evans
View
215
Download
1
Embed Size (px)
Citation preview
Slide #1 IETF 77 – Roll WG – March 2010
ROLL RPL
IETF 77 status
draft-ietf-roll-rpl
Tim WinterPascal ThubertDesign Team
RPL Status• New version draft-ietf-roll-rpl-07• DIO relatively stable• DAO refined
– for mixed source route and stateful – S flag, DTSN
• New examples in appendix B• New companion drafts:
– draft-ietf-roll-of0 – draft-levis-roll-trickle – draft-hui-6man-rpl-option
Slide #2 IETF 77 – Roll WG – March 2010
Room for simplification
• S-flag (related to #26 and LSRR)
• T-flag (vs DODAG sequence only)
• Prefix in DIOs
• Consistency (what, when)
Slide #3 IETF 77 – Roll WG – March 2010
Major Open Tickets
• #17: replacing hash with DTSN
• #21: control bits in flow label
• #23: OCP object
• #24: P2P Discussion
• #25: RPL satisfying the MUST reqs
• #26: storing / non-storing / mixed
• #27: DAO ACK
• #28: Source Route FailureSlide #4 IETF 77 – Roll WG – March 2010
Ticket #17: LSRR updates
• Problem: Updating source route path when parent moves.• Question: Do we cache source routes?
How do we trigger updates?
Cost of the update?• Proposed approach:
– DTSN triggers updates in the Source Route SUBDAG– S-flag avoids the O(Children) problem– #26 is opened to assert the need of mixed mode
• Ticket owner: Pascal Thubert
Slide #5 IETF 77 – Roll WG – March 2010
Ticket #21:Flow Label
• Problem: Data packets need an instance ID field.• Question: Relationship with flow label?• Proposed approach:
– Packets MAY have a hop-by-hop header that specifies instance ID. This header does not escape the RPL network.– If a packet does not have this hop-by-hop header, RPL assumes the instance ID is embedded in the flow label.– Gateway defaults RPL-inbound packets to instance ID 0.– Gateway decides whether to insert header based on signaling external to RPL specification.
• Ticket owner: Philip Levis
Slide #6 IETF 77 – Roll WG – March 2010
Ticket #23 OCP Object
• Question: Should RPL or Metrics draft specify the OCP object?• Background:
– Historically RPL draft did include the OCP code (as member of DIO base)– OCP object was subsequently introduced and migrated to metrics draft– As of RPL-07 inconsistent text to OCP object has been removed, leaving metrics draft to ‘own’ OCP object– RPL interface to OFs may now be left somewhat ambiguous
• Next Step: Clarify in RPL the interface to Ofs• Ticket owner: Tim Winter
Slide #7 IETF 77 – Roll WG – March 2010
Ticket #24 P2P Performance• Question: How do we make RPL really support P2P?• Background:
- Use RPL for P2P raised concerns in terms of path cost (potential issue in terms of battery)-Issue when sending traffic to nodes that do not maintain connectivity (reactiveness)
• Next Step: Use reactive DIS message across the DODAG–TTL = 5–Forward limited # of DIS copies (dampening)–Collect source route in DIS message–Option: Piggyback (short) data msg to DIS message
•Ticket owner : Anders Brandt
Slide #8 IETF 77 – Roll WG – March 2010
Ticket #25: RPL Satisfying the MUST requirements
• Question: Does RPL (rev-07) satisfy the requirements?• Proposed approach:
– List all MUST spelled out in RFC5548, RFC5673, « home routing » and « building routing » requirements documents.– Some of the MUST might now be seen as not sufficiently acurrate– Some of you mentioned that some requirements were missing (say it before !!). Still we can look at them, but won’t start again the requirements analysis.– Matrix will be built and sent to the list– Reasonable to not satisfy a MUST if clearly justified.
• Ticket owner: JP Vasseur
Slide #9 IETF 77 – Roll WG – March 2010
Ticket #26: Mixed Operation
• Problem: Mixed mode leads to complex behaviours• Question: Should RPL specify operation in either a storing or non-storing (but not mixed) mode for down routes?• Proposed approach:
– DIO control field to signal mode of operation as storing or non-storing (allow for future specification of mixed modes).– All nodes MUST support non-storing operation (source routing, LSRR or DAO?).– Nodes MAY support storing operation (hop-by-hop routing, DAO). Nodes without adequate resources participate as leaves only.– Key design point; further feedback from WG?
•Ticket owner: Tim WinterSlide #10 IETF 77 – Roll WG – March 2010
Ticket #27: DAO Ack
• Problem: DTSN covers the loss of DIO because of LLN.• Question: Should we also protect DAO?• Proposed approach:
– none yet – we are still picking advice from the list.
• Ticket owner: Pascal Thubert
Slide #11 IETF 77 – Roll WG – March 2010
Ticket 28: Source Route Failure• Problem: Source route used to support non-storing nodes• Question: What action to take when source route fails?• Approaches:
– Possible actions:• Drop packet and do nothing• Drop packet and send back ICMP error• Drop packet and send back RPL message• Backtrack by sending packet towards root
– Possible uses:• Specify only a single action to take• Allow different actions on a per-instance/dag/packet basis
• Ticket owner: Jonathan Hui
Slide #12 IETF 77 – Roll WG – March 2010
Other Room for refinement
• Generic Role of OF
• Parameters – DIO vs. OF vs. Spec vs. implementation
• Movement– When to move, jump, migrate
• DAO– When to proactively push a DAO– When to pull DAOs using DTSN and T bit.
Slide #13 IETF 77 – Roll WG – March 2010
Miscellaneous/editorial
• Reporting to the root– Eg Request a global repair
• Options in DIS– Security, filters,
• Still some inconsistencies– DAG vs. DODAG
Slide #14 IETF 77 – Roll WG – March 2010