| Network Working Group | N. Bahadur, Ed. |
| Internet-Draft | Bracket Computing |
| Intended status: Informational | R. Folkes, Ed. |
| Expires: November 28, 2014 | Juniper Networks, Inc. |
| S. Kini, Ed. | |
| Ericsson | |
| J. Medved | |
| Cisco | |
| May 27, 2014 |
Routing Information Base Info Model
draft-ietf-i2rs-rib-info-model-03
Routing and routing functions in enterprise and carrier networks are typically performed by network devices (routers and switches) using a routing information base (RIB). Protocols and configuration push data into the RIB and the RIB manager installs state into the hardware; for packet forwarding. This draft specifies an information model for the RIB to enable defining a standardized data model. Such a data model can be used to define an interface to the RIB from an entity that may even be external to the network device. This interface can be used to support new use-cases being defined by the IETF I2RS WG.
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 http://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 November 28, 2014.
Copyright (c) 2014 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 (http://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.
Routing and routing functions in enterprise and carrier networks are traditionally performed in network devices. Traditionally routers run routing protocols and the routing protocols (along with static config) populate the Routing information base (RIB) of the router. The RIB is managed by the RIB manager and the RIB manager provides a north-bound interface to its clients i.e. the routing protocols to insert routes into the RIB. The RIB manager consults the RIB and decides how to program the forwarding information base (FIB) of the hardware by interfacing with the FIB manager. The relationship between these entities is shown in Figure 1.
+-------------+ +-------------+
|RIB client 1 | ...... |RIB client N |
+-------------+ +-------------+
^ ^
| |
+----------------------+
|
V
+---------------------+
|RIB manager |
| |
| +-----+ |
| | RIB | |
| +-----+ |
+---------------------+
^
|
+---------------------------------+
| |
V V
+-------------+ +-------------+
|FIB manager 1| |FIB manager M|
| +-----+ | .......... | +-----+ |
| | FIB | | | | FIB | |
| +-----+ | | +-----+ |
+-------------+ +-------------+
Figure 1: RIB manager, RIB clients and FIB managers
Routing protocols are inherently distributed in nature and each router makes an independent decision based on the routing data received from its peers. With the advent of newer deployment paradigms and the need for specialized applications, there is an emerging need to guide the router's routing function [I-D.ietf-i2rs-problem-statement]. Traditional network-device protocol-based RIB population suffices for most use cases where distributed network control is used. However there are use cases which the network operators currently address by configuring static routes, policies and RIB import/export rules on the routers. There is also a growing list of use cases [I-D.white-i2rs-use-case], [I-D.hares-i2rs-use-case-vn-vc] in which a network operator might want to program the RIB based on data unrelated to just routing (within that network's domain). Programming the RIB could be based on other information such as routing data in the adjacent domain or the load on storage and compute in the given domain. Or it could simply be a programmatic way of creating on-demand dynamic overlays (e.g. GRE tunnels) between compute hosts (without requiring the hosts to run traditional routing protocols). If there was a standardized publicly documented programmatic interface to a RIB, it would enable further networking applications that address a variety of use-cases [I-D.ietf-i2rs-problem-statement].
A programmatic interface to the RIB involves 2 types of operations - reading from the RIB and writing (adding/modifying/deleting) to the RIB. [I-D.white-i2rs-use-case] lists various use-cases which require read and/or write manipulation of the RIB.
In order to understand what is in a router's RIB, methods like per-protocol SNMP MIBs and show output screen scraping are used. These methods are not scalable, since they are client pull mechanisms and not proactive push (from the router) mechanisms. Screen scraping is error prone (since the output format can change) and is vendor dependent. Building a RIB from per-protocol MIBs is error prone since the MIB data represent protocol data and not the exact information that went into the RIB. Thus, just getting read-only RIB information from a router is a hard task.
Adding content to the RIB from an external entity can be done today using static configuration mechanisms provided by router vendors. However the mix of what can be modified in the RIB varies from vendor to vendor and the method of configuring it is also vendor dependent. This makes it hard for an external entity to program a multi-vendor network in a consistent and vendor-independent way.
The purpose of this draft is to specify an information model for the RIB. Using the information model, one can build a detailed data model for the RIB. That data model could then be used by an external entity to program a network device.
The rest of this document is organized as follows. Section 2 goes into the details of what constitutes and can be programmed in a RIB. Guidelines for reading and writing the RIB are provided in Section 3 and Section 4 respectively. Section 5 provides a high-level view of the events and notifications going from a network device to an external entity, to update the external entity on asynchronous events. The RIB grammar is specified in Section 6. Examples of using the RIB grammar are shown in Section 7. Section 8 covers considerations for performing RIB operations at scale.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
This section describes the details of a RIB. It makes forward references to objects in the RIB grammar (Section 6). A high-level description of the RIB contents is as shown below.
routing-instance
| |
| |
0..N | | 1..N
| |
interface(s) RIB(s)
|
|
| 0..N
route(s)
Figure 2: RIB model
A RIB is an entity that contains routes. A RIB is identified by its name and a RIB is contained within a routing instance (Section 2.2). The name MUST be unique within a routing instance. All routes in a given RIB MUST be of the same type (e.g. IPv4). Each RIB MUST belong to a routing instance.
A routing instance can have multiple RIBs. A routing instance can even have two or more RIBs with the same type of routes (e.g. IPv6). A typical case where this can be used is for multi-topology routing ([RFC4915], [RFC5120]).
Each RIB can be optionally associated with a ENABLE_IP_RPF_CHECK attribute that enables Reverse path forwarding (RPF) checks on all IP routes in that RIB. Reverse path forwarding (RPF) check is used to prevent spoofing and limit malicious traffic. For IP packets, the IP source address is looked up and the rpf interface(s) associated with the route for that IP source address is found. If the incoming IP packet's interface matches one of the rpf interface(s), then the IP packet is forwarded based on its IP destination address; otherwise, the IP packet is discarded.
A routing instance, in the context of the RIB information model, is a collection of RIBs, interfaces, and routing parameters. A routing instance creates a logical slice of the router and allows different logical slices; across a set of routers; to communicate with each other. Layer 3 Virtual Private Networks (VPN), Layer 2 VPNs (L2VPN) and Virtual Private Lan Service (VPLS) can be modeled as routing instances. Note that modeling a Layer 2 VPN using a routing instance only models the Layer-3 (RIB) aspect and does not model any layer-2 information (like ARP) that might be associated with the L2VPN.
The set of interfaces indicates which interfaces are associated with this routing instance. The RIBs specify how incoming traffic is to be forwarded. And the routing parameters control the information in the RIBs. The intersection set of interfaces of 2 routing instances MUST be the null set. In other words, an interface MUST NOT be present in 2 routing instances. Thus a routing instance describes the routing information and parameters across a set of interfaces.
A routing instance MUST contain the following mandatory fields.
A routing instance MAY contain the following optional fields.
A route is essentially a match condition and an action following the match. The match condition specifies the kind of route (IPv4, MPLS, etc.) and the set of fields to match on. Figure 3 represents the overall contents of a route.
route
| | |
+---------+ | +----------+
| | |
0..N | | | 1..N
route-attribute match nexthop-list
|
|
+-------+-------+-------+--------+
| | | | |
| | | | |
IPv4 IPv6 MPLS MAC Interface
(Unicast/Multicast)
Figure 3: Route model
This document specifies the following match types:
Each route MUST have associated with it the following mandatory route attributes.
Each route can have associated with it one or more optional route attributes.
A nexthop represents an object resulting from a route lookup. For example, if a route lookup results in sending the packet out a given interface, then the nexthop represents that interface.
Nexthops can be fully resolved nexthops or unresolved nexthop. A resolved nexthop has adequate information to send the outgoing packet to the destination by forwarding it on an interface to a directly connected neighbor. For example, a nexthop to a point-to-point interface or a nexthop to an IP address on an Ethernet interface has the nexthop resolved. An unresolved nexthop is something that requires the RIB manager to determine the final resolved nexthop. For example, a nexthop could be an IP address. The RIB manager would resolve how to reach that IP address, e.g. is the IP address reachable by regular IP forwarding or by a MPLS tunnel or by both. If the RIB manager cannot resolve the nexthop, then the nexthop remains in an unresolved state and is NOT a candidate for installation in the FIB. Future RIB events can cause an unresolved nexthop to get resolved (like that IP address being advertised by an IGP neighbor). Conversely resolved nexthops can also become unresolved (e.g. in case of a tunnel going down) and hence would no longer be candidates to be installed in the FIB.
When at least one of a route's nexthops is resolved, then the route can be used to forward packets. Such a route is considered eligible to be installed in the FIB and is henceforth referred to as a FIB-eligible route. Conversely, when all the nexthops of a route are unresolved that route can no longer be used to forward packets. Such a route is considered ineligible to be installed in the FIB and is henceforth referred to as a FIB-ineligible route. The RIB information model allows an external entity to program routes whose nexthops may be unresolved initially. Whenever an unresolved nexthop gets resolved, the RIB manager will send a notification of the same (see Section 5 ).
The overall structure and usage of a nexthop is as shown in the figure below.
route
|
| 0..N
nexthop-list
|
+------------------+------------------+
1..N | |
| |
nexthop-list-member special-nexthop
|
|
nexthop-chain
|
1..N |
nexthop
|
|
+--------+------+------------------+------------------+
| | | |
| | | |
nexthop-id egress-interface logical-tunnel tunnel-encap
Figure 4: Nexthop model
Nexthops can be identified by an identifier to create a level of indirection. The identifier is set by the RIB manager and returned to the external entity on request. The RIB data-model SHOULD support a way to optionally receive a nexthop identifier for a given nexthop. For example, one can create a nexthop that points to a BGP peer. The returned nexthop identifier can then be used for programming routes to point to the same nexthop. Given that the RIB manager has created an indirection for that BGP peer using the nexthop identifier, if the transport path to the BGP peer changes, that change in path will be seamless to the external entity and all routes that point to that BGP peer will automatically start going over the new transport path. Nexthop indirection using identifiers could be applied to not just unicast nexthops, but even to nexthops that contain chains and nested nexthops (Section 2.4.1).
This document specifies a very generic, extensible and recursive grammar for nexthops. Nexthops can be Section 7.2.
It is expected that all network devices will have a limit on how many levels of lookup can be performed and not all hardware will be able to support all kinds of nexthops. RIB capability negotiation becomes very important for this reason and a RIB data-model MUST specify a way for an external entity to learn about the network device's capabilities. Examples of when and how to use various kinds of nexthops are shown in
Tunnel nexthops allow an external entity to program static tunnel headers. There can be cases where the remote tunnel end-point does not support dynamic signaling (e.g. no LDP support on a host) and in those cases the external entity might want to program the tunnel header on both ends of the tunnel. The tunnel nexthop is kept generic with specifications provided for some commonly used tunnels. It is expected that the data-model will model these tunnel types with complete accuracy.
Nexthop chains can be used to specify multiple headers over a packet, before a packet is forwarded. One simple example is that of MPLS over GRE, wherein the packet has an inner MPLS header followed by a GRE header followed by an IP header. The outermost IP header is decided by the network device whereas the MPLS header and GRE header are specified by the controller. Not every network device will be able to support all kinds of nexthop chains and an arbitrary number of header chained together. The RIB data-model SHOULD provide a way to expose nexthop chaining capability supported by a given network device.
For nexthops that are of the form of a list(s), attributes can be associated with each member of the list to indicate the role of an individual member of the list. Two kinds of attributes are specified:
A nexthop list MAY contain elements that have both PROTECTION_PREFERENCE and LOAD_BALANCE_WEIGHT set. When both are set, it means under normal operation the network device should load balance the traffic over all FIB-eligible nexthops of the current protection preference.
At the lowest level, a nexthop can be one of:
This document specifies certain special nexthops. The purpose of each of them is explained below:
A RIB data-model MUST allow an external entity to read entries, for RIBs created by that entity. The network device administrator MAY allow reading of other RIBs by an external entity through access lists on the network device. The details of access lists are outside the scope of this document.
The data-model MUST support a full read of the RIB and subsequent incremental reads of changes to the RIB. An external agent SHOULD be able to request a full read at any time in the lifecycle of the connection. When sending data to an external entity, the RIB manager SHOULD try to send all dependencies of an object prior to sending that object.
A RIB data-model MUST allow an external entity to write entries, for RIBs created by that entity. The network device administrator MAY allow writes to other RIBs by an external entity through access lists on the network device. The details of access lists are outside the scope of this document.
When writing an object to a RIB, the external entity SHOULD try to write all dependencies of the object prior to sending that object. The data-model MUST support requesting identifiers for nexthops and collecting the identifiers back in the response.
Route programming in the RIB MUST result in a return code that contains the following attributes:
The data-model MUST specify which objects are modify-able objects. A modify-able object is one whose contents can be changed without having to change objects that depend on it and without affecting any data forwarding. To change a non-modifiable object, one will need to create a new object and delete the old one. For example, routes that use a nexthop that is identified by a nexthop-identifier should be unaffected when the contents of that nexthop changes.
Asynchronous notifications are sent by the network device's RIB manager to an external entity when some event occurs on the network device. A RIB data-model MUST support sending asynchronous notifications. A brief list of suggested notifications is as below:
This section specifies the RIB information model in Routing Backus-Naur Form [RFC5511].
<routing-instance> ::= <INSTANCE_NAME>
[<interface-list>] <rib-list>
[<ROUTER_ID>]
<interface-list> ::= (<INTERFACE_IDENTIFIER> ...)
<rib-list> ::= (<rib> ...)
<rib> ::= <RIB_NAME> <rib-family>
[<route> ... ]
[ENABLE_IP_RPF_CHECK]
<rib-family> ::= <IPV4_RIB_FAMILY> | <IPV6_RIB_FAMILY> |
<MPLS_RIB_FAMILY> | <IEEE_MAC_RIB_FAMILY>
<route> ::= <match> <nexthop-list>
[<route-attributes>]
[<route-vendor-attributes>]
<match> ::= <route-type> (<ipv4-route> | <ipv6-route> | <mpls-route> |
<mac-route> | <interface-route>)
<route-type> ::= <IPV4> | <IPV6> | <MPLS> | <IEEE_MAC> | <INTERFACE>
<ipv4-route> ::= <ip-route-type>
(<destination-ipv4-address> | <source-ipv4-address> |
(<destination-ipv4-address> <source-ipv4-address>))
<destination-ipv4-address> ::= <ipv4-prefix>
<source-ipv4-address> ::= <ipv4-prefix>
<ipv4-prefix> ::= <IPV4_ADDRESS> <IPV4_PREFIX_LENGTH>
<ipv6-route> ::= <ip-route-type>
(<destination-ipv6-address> | <source-ipv6-address> |
(<destination-ipv6-address> <source-ipv6-address>))
<destination-ipv6-address> ::= <ipv6-prefix>
<source-ipv6-address> ::= <ipv6-prefix>
<ipv6-prefix> ::= <IPV6_ADDRESS> <IPV6_PREFIX_LENGTH>
<ip-route-type> ::= <SRC> | <DEST> | <DEST_SRC>
<mpls-route> ::= <MPLS_LABEL>
<mac-route> ::= <MAC_ADDRESS>
<interface-route> ::= <INTERFACE_IDENTIFIER>
<route-attributes> ::= <ROUTE_PREFERENCE> [<LOCAL_ONLY>]
[<address-family-route-attributes>]
<address-family-route-attributes> ::= <ip-route-attributes> |
<mpls-route-attributes> |
<ethernet-route-attributes>
<ip-route-attributes> ::= <>
<mpls-route-attributes> ::= <>
<ethernet-route-attributes> ::= <>
<route-vendor-attributes> ::= <>
<nexthop-list> ::= <special-nexthop> |
((<nexthop-list-member>) |
([<nexthop-list-member> ... ] <nexthop-list> ))
<nexthop-list-member> ::= (<nexthop-chain> |
<nexthop-chain-identifier> )
[<nexthop-list-member-attributes>]
<nexthop-list-member-attributes> ::= [<PROTECTION_PREFERENCE>]
[<LOAD_BALANCE_WEIGHT>]
<nexthop-chain> ::= (<nexthop> ...)
<nexthop-chain-identifier> ::= <NEXTHOP_NAME> | <NEXTHOP_ID>
<nexthop> ::= (<nexthop-identifier> | <EGRESS_INTERFACE> |
<ipv4-address> | <ipv6-address> |
(<EGRESS_INTERFACE> (<ipv4-address> | <ipv6-address>)
[RIB_NAME]) |
(<EGRESS_INTERFACE> <IEEE_MAC_ADDRESS>) |
(<tunnel-encap> [<EGRESS_INTERFACE>]) |
<logical-tunnel> |
<RIB_NAME>)
<nexthop-identifier> ::= <NEXTHOP_NAME> | <NEXTHOP_ID>
<special-nexthop> ::= <DISCARD> | <DISCARD_WITH_ERROR> |
(<RECEIVE> [<COS_VALUE>])
<logical-tunnel> ::= <tunnel-type> <TUNNEL_NAME>
<tunnel-type> ::= <IPV4> | <IPV6> | <MPLS> | <GRE> | <VxLAN> | <NVGRE>
<tunnel-encap> ::= (<IPV4> <ipv4-header>) |
(<IPV6> <ipv6-header>) |
(<MPLS> <mpls-header>) |
(<GRE> <gre-header>) |
(<VXLAN> <vxlan-header>) |
(<NVGRE> <nvgre-header>)
<ipv4-header> ::= <SOURCE_IPv4_ADDRESS> <DESTINATION_IPv4_ADDRESS>
<PROTOCOL> [<TTL>] [<DSCP>]
<ipv6-header> ::= <SOURCE_IPV6_ADDRESS> <DESTINATION_IPV6_ADDRESS>
<NEXT_HEADER> [<TRAFFIC_CLASS>]
[<FLOW_LABEL>] [<HOP_LIMIT>]
<mpls-header> ::= (<mpls-label-operation> ...)
<mpls-label-operation> ::= (<MPLS_PUSH> <MPLS_LABEL> [<S_BIT>]
[<TOS_VALUE>] [<TTL_VALUE>]) |
(<MPLS_POP> [<TTL_ACTION>])
<gre-header> ::= <GRE_IP_DESTINATION> <GRE_PROTOCOL_TYPE> [<GRE_KEY>]
<vxlan-header> ::= (<ipv4-header> | <ipv6-header>)
[<VXLAN_IDENTIFIER>]
<nvgre-header> ::= (<ipv4-header> | <ipv6-header>)
<VIRTUAL_SUBNET_ID>
[<FLOW_ID>]
Figure 5: RIB rBNF grammar
The RIB grammar is very generic and covers a variety of features. This section provides examples on using objects in the RIB grammar and examples to program certain use cases.
Using route preference a client can pre-install alternate paths in the network. For example, if OSPF has a route preference of 10, then another client can install a route with route preference of 20 to the same destination. The OSPF route will get precedence and will get installed in the FIB. When the OSPF route is withdrawn, the alternate path will get installed in the FIB.
Route preference can also be used to prevent denial of service attacks by installing routes with the best preference, which either drops the offending traffic or routes it to some monitoring/analysis station. Since the routes are installed with the best preference, they will supersede any route installed by any other protocol.
The RIB grammar allows one to create a variety of nexthops. This section describes uses for certain types of nexthops.
A tunnel nexthop points to a tunnel of some kind. Traffic that goes over the tunnel gets encapsulated with the tunnel encap. Tunnel nexthops are useful for abstracting out details of the network, by having the traffic seamlessly route between network edges.
One can create a replication list for replication traffic to multiple destinations. The destinations, in turn, could be complex nexthops in themselves - at a level supported by the network device. Point to multipoint and broadcast are examples that involve replication.
A replication list (at the simplest level) can be represented as:
<nexthop-list> ::= <nexthop> [ <nexthop> ... ] The above can be derived from the grammar as follows: <nexthop-list> ::= <nexthop-list-member> [<nexthop-list-member> ...] <nexthop-list> ::= <nexthop-chain> [<nexthop-chain> ...] <nexthop-list> ::= <nexthop> [ <nexthop> ... ]
A weighted list is used to load-balance traffic among a set of nexthops. From a modeling perspective, a weighted list is very similar to a replication list, with the difference that each member nexthop MUST have a LOAD_BALANCE_WEIGHT associated with it.
A weighted list (at the simplest level) can be represented as:
<nexthop-list> ::= (<nexthop> <LOAD_BALANCE_WEIGHT>)
[(<nexthop> <LOAD_BALANCE_WEIGHT>)... ]
The above can be derived from the grammar as follows:
<nexthop-list> ::= <nexthop-list-member> [<nexthop-list-member> ...]
<nexthop-list> ::= (<nexthop-chain> <nexthop-list-member-attributes>)
[(<nexthop-chain>
<nexthop-list-member-attributes>) ...]
<nexthop-list> ::= (<nexthop-chain> <LOAD_BALANCE_WEIGHT>)
[(<nexthop-chain> <LOAD_BALANCE_WEIGHT>) ... ]
<nexthop-list> ::= (<nexthop> <LOAD_BALANCE_WEIGHT>)
[(<nexthop> <LOAD_BALANCE_WEIGHT>)... ]
Protection lists are similar to weighted lists. A protection list specifies a set of primary nexthops and a set of backup nexthops. The <PROTECTION_PREFERENCE> attribute indicates which nexthop is primary and which is backup.
A protection list can be represented as:
<nexthop-list> ::= (<nexthop> <PROTECTION_PREFERENCE>)
[(<nexthop> <PROTECTION_PREFERENCE>)... ]
A protection list can also be a weighted list. In other words, traffic can be load-balanced among the primary nexthops of a protection list. In such a case, the list will look like:
<nexthop-list> ::= (<nexthop> <PROTECTION_PREFERENCE>
<LOAD_BALANCE_WEIGHT>)
[(<nexthop> <PROTECTION_PREFERENCE>
<LOAD_BALANCE_WEIGHT>)... ]
A nexthop chain is a nexthop that puts one or more headers on an outgoing packet. One example is a Pseudowire - which is MPLS over some transport (MPLS or GRE for instance). Another example is VxLAN over IP. A nexthop chain allows an external entity to break up the programming of the nexthop into independent pieces - one per encapsulation.
A simple example of MPLS over GRE can be represented as:
<nexthop-list> ::= (<MPLS> <mpls-header>) (<GRE> <gre-header>) The above can be derived from the grammar as follows: <nexthop-list> ::= <nexthop-list-member> [<nexthop-list-member> ...] <nexthop-list> ::= <nexthop-chain> <nexthop-list> ::= <nexthop> [ <nexthop> ... ] <nexthop-list> ::= <tunnel-encap> (<nexthop> [ <nexthop> ...]) <nexthop-list> ::= <tunnel-encap> (<tunnel-encap>) <nexthop-list> ::= (<MPLS> <mpls-header>) (<GRE> <gre-header>)
Lists of lists is a complex construct. One example of usage of such a construct is to replicate traffic to multiple destinations, with high availability. In other words, for each destination you have a primary and backup nexthop (replication list) to ensure there is no traffic drop in case of a failure. So the outer list is a protection list and the inner lists are replication lists of primary/backup nexthops.
IP multicast involves matching a packet on (S, G) or (*, G), where both S (source) and G (group) are IP prefixes. Following the match, the packet is replicated to one or more recipients. How the recipients subscribe to the multicast group is outside the scope of this document.
In PIM-based multicast, the packets are IP forwarded on an IP multicast tree. The downstream nodes on each point in the multicast tree is one or more IP addresses. These can be represented as a replication list ( Section 7.2.2 ).
In MPLS-based multicast, the packets are forwarded on a point to multipoint (P2MP) label-switched path (LSP). The nexthop for a P2MP LSP can be represented in the nexthop grammar as a <logical-tunnel> (P2MP LSP identifier) or a replication list ( Section 7.2.2) of <tunnel-encap>, with each tunnel encap representing a single mpls downstream nexthop.
This section discusses the scale requirements for a RIB data-model. The RIB data-model should be able to handle large scale of operations, to enable deployment of RIB applications in large networks.
Bulking (grouping of multiple objects in a single message) MUST be supported when a network device sends RIB data to an external entity. Similarly the data model MUST enable a RIB client to request data in bulk from a network device.
Bulking (grouping of multiple write operations in a single message) MUST be supported when an external entity wants to write to the RIB. The response from the network device MUST include a return-code for each write operation in the bulk message.
There can be cases where a single network event results in multiple events and/or notifications from the network device to an external entity. On the other hand, due to timing of multiple things happening at the same time, a network device might have to send multiple events and/or notifications to an external entity. The network device originated event/notification message MUST support bulking of multiple events and notifications in a single message.
All interactions between a RIB manager and an external entity MUST be authenticated and authorized. The RIB manager MUST protect itself against a denial of service attack by a rogue external entity, by throttling request processing. A RIB manager MUST enforce limits on how much data can be programmed by an external entity and return error when such a limit is reached.
The RIB manager MUST expose a data-model that it implements. An external agent MUST send requests to the RIB manager that comply with the supported data-model. The data-model MUST specify the behavior of the RIB manager on handling of unsupported data requests.
This document does not generate any considerations for IANA.
The authors would like to thank the working group co-chairs and reviewers on their comments and suggestions on this draft. The following people contributed to the design of the RIB model as part of the I2RS Interim meeting in April 2013 - Wes George, Chris Liljenstolpe, Jeff Tantsura, Susan Hares and Fabian Schneider.
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. |