Presentation is loading. Please wait.

Presentation is loading. Please wait.

Comparison of Link Identification schemes Objective: Present the similarities and differences of the two schemes.

Similar presentations


Presentation on theme: "Comparison of Link Identification schemes Objective: Present the similarities and differences of the two schemes."— Presentation transcript:

1 Comparison of Link Identification schemes Objective: Present the similarities and differences of the two schemes

2 AFAIK Both solutions require all the routers to learn the set of all prefixes on the link. Both solutions use the set of all prefixes implicitly as the link identifier –Primary difference is in the how it is represented in RA messages.

3 Goals G1: Link identification and validate IP configuration G2: Minimal latency G3: False change detection should be avoided. G4: No undue signaling G5: Use existing signaling mechanism

4 Goals (Contd.) G6: Use only link local scope signaling. G7: Compatible with SEND. G8: No new security vulnerabilities. G9: The nodes, such as routers or hosts, supporting DNA schemes should work appropriately with unmodified nodes, such as routers or hosts, which do not support DNA schemes. G10: Support nodes attached to multiple links.

5 FunctionLinkID draft-jinchoi-dna-protocol2 Landmark + CompleteRA draft-pentland-dna-protocol Identifier choice (1)Routers choose IDHost (Landmark) and Router choose ID (Complete Prefix list) Identifier structure(2)One prefix from LinkAt least one prefix Previous Link Prefix disclosure (3)Not required Help with complete prefix list generation (4) Rely on CPL-draft recommendation Yes Solicitation required (5)No RS required for FastRA (6)Yes DNA to DNA move (7)Single RA Non-DNA to DNA move (8)Relies on L2 HintYes DNA to non-DNA move (9)Relies on CPL –draft recommendation for CPL generation Yes: Except new router on Link Supports mixed links (10)Relies on L2 hint if the prefixes are disjoint Yes Support large # of prefixes (11)YesYes (Landmark only advantages)

6 Additional information

7 FunctionLinkIDLandmark + CompleteRA Bandwidth solicited RA (1 router) RS + RA (+ 32 if LinkID is in LPIO) (+32 for another LinkID) RS + 24 + RA + DNAO + 24: less options and DNAO if Yes answer Bandwidth unsolicited RA ( if r is the mean RA rate) r * (RA (+ 32 if LinkID is in LPIO) (+32 for another LinkID)) r * (RA + DNAO) Average solicited bandwidth (if a is APs per Link) RS + RA (+ 32 if LinkID is in LPIO) (+32 for another LinkID) RS + 24 + 1/a (RA+DNAO) + (1-1/a)(RA+24-options) Relies on CPL means: Relies on L2 Hints and whether the prefix list is actually complete. DNAO = ROOF ((17 (P – Conf) + 2)/8) * 8 Relies on Hints means: Requires L2 hints to be able to differentiate two links.

8 Footnotes 1.Identifier Choice: Indicates which device is in charge of coordinating the identifier or identifiers used for DNA. For the CompleteRA/Landmarks 2.LinkID chooses one out of all prefixes, as does landmark. CompleteRA uses all the prefixes, but the combined scheme, can use either. 3.With landmarks, the previous link’s prefix is included in the RS. It is possible in protocol-01, to send the RS without the last prefix, in which case, no landmark response is possible, but a completeRA may be sent. 4.Prefix Lists are built in draft-ietf-dna-cpl-01. Once all prefixes on the link are known, reception of RAs with no prefixes in the complete Prefix List indicate change of link. CompleteRA populates a prefix list immediately to make it complete. We assume here that Landmark/CompleteRA systems always send a completeRA if the host doesn’t already know the Complete Prefix List (if not Landmark=yes). In cases without CompleteRA, the procedures in the cpl draft must be relied upon to generate the Complete Prefix List. 5.Landmarks need a solicitation to indicate link change. CompleteRA and LinkID do not, and can just rely on comparison of received RAs to determine change

9 Footnotes 6.While this is related to the previous slide, this line was added to show that FastRA requires RS in any case. 7.Single RA refers to the fact that any single received RA can be used to determine if link change has occurred. 8.Here, ‘Relies on L2 hint’ indicates that distinguishing between a new router and a link change requires link-layer hint reception. This is the case where the new RA has no LinkID, even if the last RA did. For CompleteRA/Landmark, this assumes that the host has received a completeRA, and that any subsequent reception of an RA can be classified as a link change (if reliable L2 hints are in use, it works, as well as if there are hints from other sources such as L3 timers). 9.LinkID cannot determine immediately if there is a new router, or link change has occurred unless it already has a CPL. This also requires reliable L2 hints. CompleteRA is able to identify that change has occurred immediately, but without reliable L2 hints, may have spurious change detection when a non DNA router starts on the same link

10 Footnotes 10.RA messages from non-DNA routers become indistinguishable from link change unless reliable L2 hints are available in LinkID. CompleteRAs and Landmarks both learn about non-DNA routers’ prefixes. 11.CompleteRAs do not have an explicit bounded size, but Landmark answers can be sent even if all prefixes do not fit in one RA (To make a completeRA). The worst case scenario is if a link has one router, CompleteRA with SEND can only support 15 prefixes.

11 FunctionLandmarkCompleteRALinkIDLandmark + CompleteRA Identifier choiceHost chooses IDRouters choose ID Host chooses ID Identifier structureOne prefix from Link Set of all prefixes on Link One prefix from Link At least one prefix Previous Link Prefix disclosure RequiredNot required Complete prefix list generation Rely on CPLYesRely on CPLYes Solicitation requiredYesNo RS required for FastRA Yes DNA to DNA moveSingle RA Non-DNA to DNA move Yes: Landmark Nack Yes: CompleteRA overrides Relies on L2 HintYes DNA to non-DNA move Relies on CPLYes: Except new router on Link Relies on CPLYes: Except new router on Link Supports mixed linksYes: Landmarks even non DNA prefixes Yes: All prefixes in single message Relies on L2 hint if the prefixes are disjoint Yes Support large # of prefixes YesNo: Use CPL instead Yes

12 FunctionLandmarkCompleteRALinkIDLandmark + CompleteRA Bandwidth solicited RA (1 router) RS + 24 + RA + 24: less options if Yes answer RS + RA + DNAO RS + RA (+ 32 if LinkID is in LPIO) (+32 for another LinkID) RS + 24 + RA + DNAO + 24: less options and DNAO if Yes answer Bandwidth unsolicited RA ( if r is the mean RA rate) r * RAr * (RA + DNAO)r * (RA (+ 32 if LinkID is in LPIO) (+32 for another LinkID)) r * (RA + DNAO) Average solicited bandwidth (if a is APs per Link) RS + 24 + 1/a (RA+24) + (1- 1/a)(RA+24-options) RS + RA + DNAO RS + RA (+ 32 if LinkID is in LPIO) (+32 for another LinkID) RS + 24 + 1/a (RA+DNAO) + (1- 1/a)(RA+24-options) Relies on CPL means: Relies on L2 Hints and whether the prefix list is actually complete. DNAO = ROOF ((17 (P – Conf) + 2)/8) * 8 Relies on Hints means: Requires L2 hints to be able to differentiate two links.


Download ppt "Comparison of Link Identification schemes Objective: Present the similarities and differences of the two schemes."

Similar presentations


Ads by Google