TWAMP Features – Reflect OCTETS draft draft-ietf-ippm-reflect-octets-01 Al Morton and Len Ciavattone March, 2009

2 2009, The year of New TWAMP Features  RFC 5357  Lots of implementations  New readers = New Ideas  IPPM WG added this new feature to the charter in 2008

3 Refresh: TWAMP Entities controller responder | Control-Client | | Server | | | | | | Session-Sender | | Session-Reflector |

4 Reflect Octets in BOTH Control and Test Packets Reflect Octets in BOTH Control and Test Packets  TWAMP-Control: –Contol-Client Inserts 2 octets in Request-TW-Session –Server moves 2 octets from Request- TW-Session message to the Accept (reply) message, and –Can Insert 2 Octets of its own, and –Learns the length (N) of padding in Test packets that it will need to reflect. –Length N < Overall Padding Length

5 Reflect Octets in BOTH Control and Test Packets (2) Reflect Octets in BOTH Control and Test Packets (2)  TWAMP-Test: –Session-Reflector moves N octets of the Padding to the reflected Test packet  Control-Client and Session-Sender know what was put in, the format used, etc.  Adds some possibilities for senders

6 Truncate Padding Option Truncate Padding Option  TWAMP Test packet formats are: – octets from Sender to Reflector – octets from Reflector to Sender  Most want symmetrical sizes, so…  TWAMP core spec RECOMMENDS –Reflector reduce pad. to compensate –Sender SHOULD add sufficient padding  This feature makes the behavior above a selectable option (certain).  Interaction w/ Reflect Octets Option

7 Reflect Octets: New Modes Field Values Value Description Reference/Explanation Unauth. TEST protocol, new bit position (3) 8 Unauth. TEST protocol, new bit position (3) Encrypted CONTROL Encrypted CONTROL xxx Reflect Octets new bit position (X) xxx Reflect Octets new bit position (X) Capability Capability yyy Truncate Padding new bit position (Y) yyy Truncate Padding new bit position (Y) Capability Capability

8 TWAMP Control: Request-TW-Session Command | 5 | MBZ | IPVN | Conf-Sender | Conf-Receiver | | Number of Schedule Slots | Many fields not shown | Padding Length | Two fields not shown... | | Type-P Descriptor | | Octets to be reflected | Length of padding to reflect | was MBZ | MBZ (4 octets) | | | HMAC (16 octets) | |

9 TWAMP Control: Accept-Session Command | Accept | MBZ | Port | | | SID (16 octets) | | | Reflected octets | Server Octets | | MBZ (8 octets) | | | | HMAC (16 octets) | |

10 Reflect Packet Padding: TWAMP Test Sender packet | Sequence Number | | Timestamp | | | Error Estimate | MBZ (2 octets) | | | Packet Padding (to be reflected) |. (length in octets specified elsewhere) Additional Packet Padding

11 Reflect Octets: TWAMP Test Reflector packet | Sequence Number | | Timestamp | | | | Error Estimate | MBZ | | Receive Timestamp | | | | Sender Sequence Number | | Sender Timestamp | | | | Sender TTL | Packet Padding (from Session-Sender) | Packet Padding (from Session-Sender) | | | Additional Packet Padding

12 Question for IPPM Would a minimum re-write, no truncation format be desirable? Would a minimum re-write, no truncation format be desirable? Server Info Fields Sender -> Reflector Padding to be Reflected Additional Padding Reflector Info Fields Reflector -> Sender Reflected Padding Additional Padding Truncated Padding Snd TTL Server Info Fields

13 Question for IPPM (2) OR, Should we deal with the un-even boundaries and the truncation issue using NEW Test Packet Formats that specify the Reflected Padding? OR, Should we deal with the un-even boundaries and the truncation issue using NEW Test Packet Formats that specify the Reflected Padding? Sender -> Reflector Padding to be Reflected Additional Padding Reflector -> Sender Reflected Padding Additional Padding Fill for Discard Snd TTL MBZ Server Info Fields Reflector Info Fields Server Info Fields TLV Format ?

14 Next Steps  We need a conclusive discussion before major surgery on the draft to try new Test Formats  You can appreciate some of the simplification possible IF you read the current draft (length rules).  Comments on all aspects appreciated, Authors are fully satisfied with the TWAMP-Control modifications…