| ROLL | S. Anamalamudi |
| Internet-Draft | Huaiyin Institute of Technology |
| Intended status: Standards Track | M. Zhang |
| Expires: September 6, 2018 | Huawei Technologies |
| AR. Sangi | |
| Huaiyin Institute of Technology | |
| C. Perkins | |
| Futurewei | |
| S.V.R.Anand | |
| Indian Institute of Science | |
| B. Liu | |
| Huawei Technologies | |
| March 5, 2018 |
Asymmetric AODV-P2P-RPL in Low-Power and Lossy Networks (LLNs)
draft-ietf-roll-aodv-rpl-03
Route discovery for symmetric and asymmetric Point-to-Point (P2P) traffic flows is a desirable feature in Low power and Lossy Networks (LLNs). For that purpose, this document specifies a reactive P2P route discovery mechanism for both hop-by-hop routing and source routing: Ad Hoc On-demand Distance Vector Routing (AODV) based RPL protocol. Paired Instances are used to construct directional paths, in case some of the links between source and target node are asymmetric.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 6, 2018.
Copyright (c) 2018 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
RPL[RFC6550] is a IPv6 distance vector routing protocol for Low-power and Lossy Networks (LLNs), and is designed to support multiple traffic flows through a root-based Destination-Oriented Directed Acyclic Graph (DODAG). Typically, a router does not have routing information for most other routers. Consequently, for traffic between routers within the DODAG (i.e., Point-to-Point (P2P) traffic) data packets either have to traverse the root in non-storing mode, or traverse a common ancestor in storing mode. Such P2P traffic is thereby likely to traverse sub-optimal routes and may suffer severe congestion near the DAG root [RFC6997], [RFC6998].
To discover optimal paths for P2P traffic flows in RPL, P2P-RPL [RFC6997] specifies a temporary DODAG where the source acts as a temporary root. The source initiates DIOs encapsulating the P2P Route Discovery option (P2P-RDO) with an address vector for both hop-by-hop mode (H=1) and source routing mode (H=0). Subsequently, each intermediate router adds its IP address and multicasts the P2P mode DIOs, until the message reaches the target node (TargNode). TargNode sends the "Discovery Reply" object. P2P-RPL is efficient for source routing, but much less efficient for hop-by-hop routing due to the extra address vector overhead. However, for symmetric links, when the P2P mode DIO message is being multicast from the source hop-by-hop, receiving nodes can infer a next hop towards the source. When TargNode subsequently replies to the source along the established forward route, receiving nodes determine the next hop towards TargNode. In other words, it is efficient to use only routing tables for P2P-RDO message instead of "Address vector" for hop-by-hop routes (H=1) over symmetric links.
RPL and P2P-RPL both specify the use of a single DODAG in networks of symmetric links, where the two directions of a link MUST both satisfy the constraints of the objective function. This eliminates the possibility to use asymmetric links which are qualified in one direction. But, application-specific routing requirements as defined in IETF ROLL Working Group [RFC5548], [RFC5673], [RFC5826] and [RFC5867] may be satisfied by routing paths using bidirectional asymmetric links. For this purpose, [I-D.thubert-roll-asymlink] describes bidirectional asymmetric links for RPL [RFC6550] with Paired DODAGs, for which the DAG root (DODAGID) is common for two Instances. This can satisfy application-specific routing requirements for bidirectional asymmetric links in core RPL [RFC6550]. Using P2P-RPL twice with Paired DODAGs, on the other hand, requires two roots: one for the source and another for the target node due to temporary DODAG formation. For networks composed of bidirectional asymmetric links (see Section 5), AODV-RPL specifies P2P route discovery, utilizing RPL with a new MoP. AODV-RPL makes use of two multicast messages to discover possibly asymmetric routes, which can achieve higher route diversity. AODV-RPL eliminates the need for address vector control overhead in hop-by-hop mode. This significantly reduces the control packet size, which is important for Constrained LLN networks. Both discovered routes (upward and downward) meet the application specific metrics and constraints that are defined in the Objective Function for each Instance [RFC6552].
The route discovery process in AODV-RPL is modeled on the analogous procedure specified in AODV [RFC3561]. The on-demand nature of AODV route discovery is natural for the needs of peer-to-peer routing in RPL-based LLNs. AODV terminology has been adapted for use with AODV-RPL messages, namely RREQ for Route Request, and RREP for Route Reply. AODV-RPL currently omits some features compared to AODV -- in particular, flagging Route Errors, blacklisting unidirectional links, multihoming, and handling unnumbered interfaces.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119]. Additionally, this document uses the following terms:
With AODV-RPL, routes from OrigNode to TargNode within the LLN network established are "on-demand". In other words, the route discovery mechanism in AODV-RPL is invoked reactively when OrigNode has data for delivery to the TargNode but existing routes do not satisfy the application's requirements. The routes discovered by AODV-RPL are point-to-point; in other words the routes are not constrained to traverse a common ancestor. Unlike core RPL [RFC6550] and P2P-RPL [RFC6997], AODV-RPL can enable asymmetric communication paths in networks with bidirectional asymmetric links. For this purpose, AODV-RPL enables discovery of two routes: namely, one from OrigNode to TargNode, and another from TargNode to OrigNode. When possible, AODV-RPL also enables symmetric route discovery along Paired DODAGs (see Section 5).
In AODV-RPL, route discovery is initiated by forming a temporary DAG rooted at the OrigNode. Paired DODAGs (Instances) are constructed according to a new AODV-RPL Mode of Operation (MoP) during route formation between the OrigNode and TargNode. The RREQ-Instance is formed by route control messages from OrigNode to TargNode whereas the RREP-Instance is formed by route control messages from TargNode to OrigNode (as shown in Figure 4). Intermediate routers join the Paired DODAGs based on the rank as calculated from the DIO message. Henceforth in this document, the RREQ-DIO message means the AODV-RPL mode DIO message from OrigNode to TargNode, containing the RREQ option. Similarly, the RREP-DIO message means the AODV-RPL mode DIO message from TargNode to OrigNode, containing the RREP option. Subsequently, the route discovered in the RREQ-Instance is used for data transmission from TargNode to OrigNode, and the route discovered in RREP-Instance is used for Data transmission from OrigNode to TargNode.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type | Option Length |S|H|X| Compr | L | MaxRank | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Orig SeqNo | | +-+-+-+-+-+-+-+-+ | | | | | | Address Vector (Optional, Variable Length) | | | | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 1: DIO RREQ option format for AODV-RPL MoP
A RREQ-DIO message MUST carry exactly one RREQ option.
A RREP-DIO message MUST carry exactly one RREP option.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Type | Option Length |H|X| Compr | L | MaxRank |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|T|G| SHIFT | Reserved | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| |
| |
| Address Vector (Optional, Variable Length) |
. .
. .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 2: DIO RREP option format for AODV-RPL MoP
The TargNode supplies the following information in the RREP option:
The AODV-RPL Target Option is defined based on the Target Option in core RPL [RFC6550]: the Destination Sequence Number of the TargNode is added.
A RREQ-DIO message MUST carry at least one AODV-RPL Target Options. A RREP-DIO message MUST carry exactly one AODV-RPL Target Option encapsulating the address of the OrigNode if the 'T' bit is set to 1.
If an OrigNode want to discover routes to multiple TargNodes, and these routes share the same constraints, then the OrigNode can include all the addresses of the TargNodes into multiple AODV-RPL Target Options in the RREQ-DIO, so that the cost can be reduced to building only one DODAG. Different addresses of the TargNodes can merge if they share the same prefix.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Type | Option Length | Dest SeqNo | Prefix Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ |
| Target Prefix (Variable Length) |
. .
. .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 3: Target option format for AODV-RPL MoP
In Figure 4 and Figure 5, BR is the BorderRouter, O is the OrigNode, R is an intermediate router, and T is the TargNode. If the RREQ-DIO arrives over an interface that is known to be symmetric, and the 'S' bit is set to 1, then it remains as 1, as illustrated in Figure 4. An intermediate router sends out RREQ-DIO with the 'S' bit set to 1, meaning that all the one-hop links on the route from the OrigNode to this router meet the requirements of route discovery; thus the route can be used symmetrically.
BR
/ | \
/ | \
/ | \
R R R
/ \ | / \
/ \ | / \
/ \ | / \
R -------- R --- R ----- R -------- R
/ \ <--S=1--> / \ <--S=1--> / \
<--S=1--> \ / \ / <--S=1-->
/ \ / \ / \
O ---------- R ------ R------ R ----- R ----------- T
/ \ / \ / \ / \
/ \ / \ / \ / \
/ \ / \ / \ / \
R ----- R ----------- R ----- R ----- R ----- R ---- R----- R
>---- RREQ-Instance (Control: S-->D; Data: D-->S) ------->
<---- RREP-Instance (Control: D-->S; Data: S-->D) -------<
Figure 4: AODV-RPL with Symmetric Paired Instances
Upon receiving a RREQ-DIO with the 'S' bit set to 1, a node MUST decide if this one-hop link can be used symmetrically, i.e., both the two directions meet the requirements of data transmission. If the RREQ-DIO arrives over an interface that is not known to be symmetric, or is known to be asymmetric, the 'S' bit is set to 0. Moreover, if the 'S' bit arrives already set to be '0', it is set to be '0' on retransmission (Figure 5). Therefore, for asymmetric route, there is at least one hop which doesn't fulfill the constraints in the two directions. Based on the 'S' bit received in RREQ-DIO, the TargNode decides whether or not the route is symmetric before transmitting the RREP-DIO message upstream towards the OrigNode.
The criterion and the corresponding metric used to determine if a one-hop link is symmetric or not is implementation specific and beyond the scope of the document. Also, the difference in the metric values for upward and downward directions of a link that can be establish its symmetric and asymmetric nature is implementation specific. For instance, the intermediate routers MAY choose to use local information (e.g., bit rate, bandwidth, number of cells used in 6tisch), a priori knowledge (e.g. link quality according to previous communication) or estimate the metric using averaging techniques or any other means that is appropriate to the application context.
BR
/ | \
/ | \
/ | \
R R R
/ \ | / \
/ \ | / \
/ \ | / \
R --------- R --- R ---- R --------- R
/ \ --S=1--> / \ --S=0--> / \
--S=1--> \ / \ / --S=0-->
/ \ / \ / \
O ---------- R ------ R------ R ----- R ----------- T
/ \ / \ / \ / \
/ <--S=0-- / \ / \ / <--S=0--
/ \ / \ / \ / \
R ----- R ----------- R ----- R ----- R ----- R ---- R----- R
<--S=0-- <--S=0-- <--S=0-- <--S=0-- <--S=0--
>---- RREQ-Instance (Control: S-->D; Data: D-->S) ------->
<---- RREP-Instance (Control: D-->S; Data: S-->D) -------<
Figure 5: AODV-RPL with Asymmetric Paired Instances
Appendix A describes an example method using the ETX and RSSI to estimate whether the link is symmetric in terms of link quality is given in using an averaging technique.
The route discovery process is initiated on-demand when an application at the OrigNode has data to be transmitted to the TargNode, but no route for the target exists or the current routes don't fulfill the requirements of the data transmission. In this case, the OrigNode MUST build a local RPLInstance and a DODAG rooted at itself. Then it begins to send out DIO message in AODV-RPL MoP via link-local multicast. The DIO MUST contain exactly one RREQ option as defined in Section 4.1, and at least one AODV-RPL Target Option as defined in Figure 3. This DIO message is noted as RREQ-DIO. The 'S' bit in RREQ-DIO sent out by the OrigNode is set as 1.
The maintenance of Originator and Destination Sequence Number in the RREQ option is as defined in AODV [RFC3561].
The address in the AODV-RPL Target Option can be a unicast IPv6 address, a prefix or a multicast address. The OrigNode can initiate the route discovery process for multiple targets simultaneously by including multiple AODV-RPL Target Options, and within a RREQ-DIO the requirements for the routes to different TargNodes MUST be the same.
The OrigNode can maintain different RPLInstances to discover routes with different requirements to the same targets. Due to the InstanceID pairing mechanism Section 6.3.3, route replies (RREP-DIOs) from different paired RPLInstances can be distinguished.
The transmission of RREQ-DIO follows the Trickle timer. When the L duration has transpired, the OrigNode MUST leave the DODAG and stop sending any RREQ-DIOs in the related RPLInstance.
Upon receiving a RREQ-DIO, a router out of the RREQ-instance goes through the following steps:
When a RREQ-DIO arrives at a TargNode with the 'S' bit set to 1, it means there exists a symmetric route in which the two directions can fulfill the requirements. Other RREQ-DIOs can bring the upward direction of asymmetric routes (i.e. S=0). How to choose between a qualified symmetric route and an asymmetric route hopefully having better performance is implementation-specific and out of scope. If the implementation choose to use the symmetric route, the TargNode MAY send out the RREP-DIO after a duration RREP_WAIT_TIME to wait for the convergence of RD to an optimal symmetric route.
For symmetric route, the RREP-DIO message is sent via unicast to the OrigNode; therefore the DODAG in RREP-Instance doesn't need to be actually built. The RPLInstanceID in the RREP-Instance is paired as defined in Section 6.3.3. The 'S' bit in the base DIO remains as 1. In the RREP option, The 'SHIFT' field and the 'T' bit are set as defined in Section 6.3.3. The address vector received in the RREQ-DIO MUST be included in this RREP option in case the 'H' bit is set to 0 (both in RREQ-DIO and RREP-DIO). If the 'T' bit is set to 1, the address of the OrigNode MUST be encapsulated in an AODV-RPL Target Option and included in this RREP-DIO message, and the Destination Sequence Number is set according to AODV [RFC3561].
When a RREQ-DIO arrives at a TargNode with the 'S' bit set to 0, the TargNode MUST build a DODAG in the RREP-Instance rooted at itself in order to discover the downstream route from the OrigNode to the TargNode. The RREP-DIO message MUST be send out via link-local multicast until the OrigNode is reached or the MaxRank limit is exceeded.
The settings of the RREP-DIO are the same as in symmetric route.
Since the RPLInstanceID is assigned locally (i.e., there is no coordination between routers in the assignment of RPLInstanceID) the tuple (RPLInstanceID, DODAGID, Address in the AODV-RPL Target Option) is needed to uniquely identify a DODAG in an AODV-RPL instance. Between the OrigNode and the TargNode, there can be multiple AODV-RPL instances when applications upper layer have different requirements. Therefore the RREQ-Instance and the RREP-Instance in the same route discovery MUST be paired. The way to realize this is to pair their RPLInstance IDs.
0 1 2 3 4 5 6 7
+-+-+-+-+-+-+-+-+
|1|D| ID | Local RPLInstanceID in 0..63
+-+-+-+-+-+-+-+-+
Figure 6: Local Instance ID
Typically, the two InstanceIDs are set as the local InstanceID in core RPL:
The first bit is set to 1 indicating the RPLInstanceID is local. The 'D' bit here is used to distinguish the two AODV-RPL instances: D=0 for RREQ-Instance, D=1 for RREP-Instance. The ID of 6 bits SHOULD be the same for RREQ-Instance and RREP-Instance. Here, the 'D' bit is used slightly differently than in RPL.
When preparing the RREP-DIO, a TargNode could find the RPLInstanceID to be used for the RREP-Instance is already occupied by another instance from an earlier route discovery operation which is still active. In other words, two OrigNodes need routes to the same TargNode and they happen to use the same RPLInstanceID for RREQ-Instance. In this case, the occupied RPLInstanceID MUST NOT be used again. Then this RPLInstanceID SHOULD be shifted into another integer and shifted back to the original one at the OrigNode. In RREP option, the SHIFT field indicates the how many the original RPLInstanceID is shifted. When the new InstanceID after shifting exceeds 63, it will come back counting from 0. For example, the original InstanceID is 60, and shifted by 6, the new InstanceID will be 2. The 'T' MUST be set to 1 to make sure the two RREP-DIOs can be distinguished by the address of the OrigNode in the AODV-RPL Target Option.
Upon receiving a RREP-DIO, a router out of the RREP-Instance goes through the following steps:
In some cases, an Intermediate router that receives a RREQ-DIO message MAY transmit a "Gratuitous" RREP-DIO message back to OrigNode instead of continuing to multicast the RREQ-DIO towards TargNode. The intermediate router effectively builds the RREP-Instance on behalf of the actual TargNode. The 'G' bit of the RREP option is provided to distinguish the Gratuitous RREP-DIO (G=1) sent by the Intermediate node from the RREP-DIO sent by TargNode (G=0).
The gratuitous RREP-DIO can be sent out when an intermediate router R receives a RREQ-DIO for a TargNode T, and R happens to have both forward and reverse routes to T which also fulfill the requirements.
In case of source routing, the intermediate router R MUST unicast the received RREQ-DIO to TargNode T including the address vector between the OrigNode O and the router R. Thus T can have a complete address vector between O and itself. Then T MUST unicast a RREP-DIO including the address vector between T and R.
In case of hop-by-hop routing, R MUST unicast the received RREQ-DIO to T. The routers along the route SHOULD build new route entries with the related RPLInstanceID and DODAGID in the downward direction. Then T MUST unicast the RREP-DIO to R, and the routers along the route SHOULD build new route entries in the upward direction. Upon received the unicast RREP-DIO, R sends the gratuitous RREP-DIO to the OrigNode as the same way defined in Section 6.3.
The trickle timer operation to control RREQ-Instance/RREP-Instance multicast is similar to that in P2P-RPL [RFC6997].
+-------------+---------------+---------------+
| Value | Description | Reference |
+-------------+---------------+---------------+
| TBD1 (5) | AODV-RPL | This document |
+-------------+---------------+---------------+
Figure 7: Mode of Operation
IANA is required to assign a new Mode of Operation, named "AODV-RPL" for Point-to-Point(P2P) hop-by-hop routing under the RPL registry. The value of TBD1 is assigned from the "Mode of Operation" space [RFC6550].
+-------------+------------------------+---------------+
| Value | Meaning | Reference |
+-------------+------------------------+---------------+
| TBD2 (0x0A) | RREQ Option | This document |
+-------------+------------------------+---------------+
| TBD3 (0x0B) | RREP Option | This document |
+-------------+------------------------+---------------+
| TBD3 (0x0C) | AODV-RPL Target Option | This document |
+-------------+------------------------+---------------+
Figure 8: AODV-RPL Options
Three entries are required for new AODV-RPL options "RREQ", "RREP" and "AODV-RPL Target" with values of TBD2 (0x0A), TBD3 (0x0B) and TBD4 (0x0C) from the "RPL Control Message Options" space [RFC6550].
This document does not introduce additional security issues compared to base RPL. For general RPL security considerations, see [RFC6550].
There has been some discussion about how to determine the initial state of a link after an AODV-RPL-based network has begun operation. The current draft operates as if the links are symmetric until additional metric information is collected. The means for making link metric information is considered out of scope for AODV-RPL. In the future, RREQ and RREP messages could be equipped with new fields for use in verifying link metrics. In particular, it is possible to identify unidirectional links; an RREQ received across a unidirectional link has to be dropped, since the destination node cannot make use of the received DODAG to route packets back to the source node that originated the route discovery operation. This is roughly the same as considering a unidirectional link to present an infinite cost metric that automatically disqualifies it for use in the reverse direction.
| [I-D.thubert-roll-asymlink] | Thubert, P., "RPL adaptation for asymmetrical links", Internet-Draft draft-thubert-roll-asymlink-02, December 2011. |
Source---------->NodeA---------->NodeB------->Destination
Figure 9: Communication link from Source to Destination
We have tested the combination of "RSSI(downstream)" and "ETX (upstream)" to decide whether the link is symmetric or asymmetric at the intermediate nodes. The example of how the ETX and RSSI values are used in conjuction is explained below:
| RSSI at NodeA for NodeB | Expected ETX at NodeA for NodeB->NodeA |
|---|---|
| > -15 | 150 |
| -25 to -15 | 192 |
| -35 to -25 | 226 |
| -45 to -35 | 662 |
| -55 to -45 | 993 |
We tested the operations in this specification by making the following experiment, using the above parameters. In our experiment, a communication link is considered as symmetric if the ETX value of NodeA->NodeB and NodeB->NodeA (See Figure.8) are, say, within 1:3 ratio. This ratio should be taken as a notional metric for deciding link symmetric/asymmetric nature, and precise definition of the ratio is beyond the scope of the draft. In general, NodeA can only know the ETX value in the direction of NodeA -> NodeB but it has no direct way of knowing the value of ETX from NodeB->NodeA. Using physical testbed experiments and realistic wireless channel propagation models, one can determine a relationship between RSSI and ETX representable as an expression or a mapping table. Such a relationship in turn can be used to estimate ETX value at nodeA for link NodeB--->NodeA from the received RSSI from NodeB. Whenever nodeA determines that the link towards the nodeB is bi-directional asymmetric then the "S" bit is set to "S=0". Later on, the link from NodeA to Destination is asymmetric with "S" bit remains to "0".