]> China Telecom

Beiqijia Town, Changping District Beijing Beijing 102209 China weiwang94@foxmail.com

China Telecom

Beiqijia Town, Changping District Beijing Beijing 102209 China wangaj3@chinatelecom.cn

Huawei Technologies

Huawei Building, No.156 Beiqing Rd. Beijing Beijing 100095 China rainsword.wang@huawei.com

Verizon Inc.

13101 Columbia Pike Silver Spring MD 20904 United States of America 301 502-1347 gyan.s.mishra@verizon.com

Huawei Technologies

Huawei Building, No.156 Beiqing Rd. Beijing Beijing 100095 China jie.dong@huawei.com

RTG Area IDR Working Group RFC This draft defines a new Outbound Route Filter (ORF) type, called the VPN Prefix ORF. The described VPN Prefix ORF mechanism is applicable when the VPN routes from different VRFs are exchanged via one shared BGP session.

analyzed the scenarios and requirements for VPN routes control in the shared BGP session. This draft analyzes the existing solutions and their limitations for these scenarios, proposes the new VPN Prefix ORF solution to meet the requirements that described in section 8 of . Now, there are several solutions can be used to alleviate these problem: Route Target Constraint (RTC) as defined in Address Prefix ORF as defined in CP-ORF mechanism as defined in PE-CE edge peer Maximum Prefix Configure the Maximum Prefix for each VRF on edge nodes However, there are limitations to existing solutions: 1) Route Target Constraint RTC can only filter the VPN routes from the uninterested VRFs, if the “offending routes” come from the interested VRF, RFC mechanism can't filter them. 2) Address Prefix ORF Using Address Prefix ORF to filter VPN routes need to pre-configuration, but it is impossible to know which prefix may cause overflow in advance. 3) CP-ORF Mechanism defines the Covering Prefixes ORF (CP-ORF). A BGP speaker sends a CP-ORF to a peer in order to pull routes that cover a specified host address. A prefix covers a host address if it can be used to forward traffic towards that host address. CP-ORF is applicable in Virtual Hub-and-Spoke VPN and also the BGP/MPLS Ethernet VPN (EVPN) networks, but its main aim is also to get the wanted VPN prefixes and can't be used to filter the overwhelmed VPN prefixes dynamically. 4) PE-CE edge peer Maximum Prefix The BGP Maximum-Prefix feature is used to control how many prefixes can be received from a neighbor. By default, this feature allows a router to bring down a peer when the number of received prefixes from that peer exceeds the configured Maximum-Prefix limit. This feature is commonly used for external BGP peers. If it is applied to internal BGP peers, for example the VPN scenarios, all the VPN routes from different VRFs will share the common fate, which is not desirable for the finer control of the VPN Prefixes advertisement. 5) Configure the Maximum Prefix for each VRF on edge nodes When a VRF overflows, it stops the import of routes and log the extra VPN routes into its RIB. However, PEs still need to parse the BGP updates. These processes will cost CPU cycles and further burden the overflowed PE. This draft defines a new ORF-type, called the VPN Prefix ORF. This mechanism is event-driven and does not need to be pre-configured. When a VRF of a router overflows, the router will find out the VPN prefix (RD, RT, source PE, etc.) of offending VPN routes in this VRF, and send a VPN Prefix ORF to its BGP peer that carries the relevant information. If a BGP speaker receives a VPN Prefix ORF entry from its BGP peer, it will filter the VPN routes it tends to send according to the entry. The purpose of this mechanism is to control the outrage within the minimum range and avoid churn effects when a VRF on a device in the network overflows. VPN Prefix ORF is applicable when the VPN routes from different VRFs are exchanged via one shared BGP session.

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 .

The following terms are defined in this draft: RD: Route Distinguisher, defined in ORF: Outbound Route Filter, defined in AFI: Address Family Identifier, defined in SAFI: Subsequent Address Family Identifier, defined in EVPN: BGP/MPLS Ethernet VPN, defined in RR: Router Reflector, provides a simple solution to the problem of IBGP full mesh connection in large-scale IBGP implementation. VRF: Virtual Routing Forwarding, a virtual routing table based on VPN instance.

The operation of VPN Prefix ORF mechanism on each device is independent, each of them makes a local judgement to determine whether it needs to send VPN Prefix ORF to its upstream peer. Operators can set the algorithms in the devices according to their own circumstances. The general procedures of the VPN Prefix ORF mechanism on the sender are as follows: This section describes the procedures for the BGP peer(the first BGP peer) receiving the VPN routes information from another BGP peer(the second BGP peer). The VPN information includes the updated VPN routes and their corresponding VPN instance identification information. Based on the VPN instance identification information, the first BGP peer determines the newly added VPN routes, and checks whether the number of the newly added VPN route has caused the number of total VPN routes exceeding the maximum route limit of the associated VPN instance. If the route number of the VPN instance that identified by the VPN instance identification information is reached or excess its limit, it will send the instruction information to the second BGP peer, indicating that the second BGP peer stop sending the corresponding VPN routes that identified by the VPN instance identification information. The first BGP peer and the second BGP peer are iBGP peers that are within one AS. The VPN instance identification information is RD and the instruction information is sent via the route-refresh message. The instruction information that sends to the second BGP peer includes the followings information: The ORF entries that are included in the route-refresh message. Set the Action field in the ORF entries to the value that instructs to add the corresponding filter condition into outbound route filter of the second the BGP peer. Set the Match field in the ORF entries to the value that instructs to deny the VPN routes updates that matches the corresponding ORF entries. The RD value that identifies the above mentioned VPN instance is added at the type-specific part of the ORF entries. When multiple VRFs on a PE are receiving VPN routes for a specific RD, if one VRF exceeds its limit upon receiving routes with that RD and the PE sends a VP-ORF message, it will prevent other VRFs that have not exceeded their limits from receiving VPN routes containing that RD, leading to communication disruptions between these VRFs and the rejected VPN routes. In order to more finely control VPN routing, when the above situation occurs, VRF should generate a filter when its route limit is exceeded, to selectively filter route entries based on RD before they enter the VRF from the RIB. The PE which receives the offending VPN routes parses the received BGP update message to obtain routing information, and obtains VPN routing entries associated with the BGP path from the routing information. Then, PE should determine the target VRF based on the RT import information of the routing target entry, and configure filters for the target VRF. PE should first detect whether there is currently a filter associated with the target VRF. If yes, PE should use the filter to filter the VPN routing entries, so that VPN routing entries carrying the specified routing identifier RD are not introduced into the target VRF. If there is currently no filter associated with the target VRF, PE will directly introduce the VPN routing entry into the target VRF. The format of VPN Prefix ORF entry is shown in When configuring a filter for the target VRF, the PE should detect whether there is currently a routing overflow issue with the target VRF. If the target VRF currently has a routing overflow issue and the reason for the routing overrun is caused by a VPN routing entry carrying the specified RD, a filter is generated for the target VRF, wherein the filter is used to filter VPN routing entries carrying the specified RD. If the target VRF currently does not have routing overflow issues, PE will delete the filter for the target VRF. The detail procedures for further subdivisions are described below: a) No quota value is set on PE On PE, each VRF has a prefix limit. When the PE receives VPN routes from its BGP peer, due to the received VPN routes may belong to different VPN and carry the corresponding RDs, the PE should extract the VPN route information from BGP UPDATE message which contains VPN routes related to BGP optimal routing. PE can determine the target VRFs of the received VPN route based on the RT of the VPN route and the RT-import of VRFs. Then, the PE should sequentially determine whether each target VRF will exceed the limit after importing the received VPN routes. If a target VRF exceeds the limit which is caused by the VPN routes carrying a certain RD and the other target VRFs have not overflow, PE should not trigger the VPN Prefix ORF mechanism, and only performs VPN route filtering for the target VRF and stop importing VPN routes carrying the specific RD. If a target VRF exceeds the limit and there is no other VRFs need these VPN routes, the PE should trigger the VPN Prefix ORF mechanism and send a BGP ROUTE-REFRESH message contains the corresponding VPN Prefix ORF entry to its peer, which will generate a VPN routes filtering strategy for the VRF. And if the "Offending VPN routes process method" bit is 1, the receiver of VPN Prefix ORF entry should withdraw the extra VPN routes according to the value of VRF Prefix Limit, RD, RT and information in optional TLVs in the entry, and stop sending the corresponding VPN routes to the sender. If the target VRF no longer exceeds the limit, the relevant VPN routing filtering strategy needs to be deleted. When importing VPN routes to a VRF, it is necessary to determine whether there is a VPN routes filtering strategy on the PE for that VRF. BGP peers supporting this functionality has VRFs that has maximum route limit reached should comply with the procedures defined in this document b) Quota value is set on PE On PE, each VRF has a prefix limit, and routes associated with each <RD, source PE> tuple has a pre-configured quota. Considering that the received VPN routes may belong to different VPNs and therefore carry different RDs, the PE should determine whether VRF will exceed the limit after adding the received VPN routes. when routes associated with <RD, source PE> tuple pass the quota but the prefix limit of VRF is not exceeded, PE should send warnings to the operator, and the VPN Prefix ORF mechanism should not be triggered. when routes associated with <RD, source PE> tuple pass the quota and the prefix limit is exceeded and there is no other VRFs on offended PE need VPN routes with this <RD, source PE> tuple, PE should trigger VPN Prefix ORF mechanism and send a BGP ROUTE-REFRESH message contains the corresponding VPN Prefix ORF entry to its peer. When the VPN Prefix ORF mechanism is triggered, the device must send an alarm information to network operators.

For intra-AS VPN deployment, there are three scenarios: RD is allocated per VPN per PE, each VRF only import one RT (see ). RD is allocated per VPN per PE. Multiple RTs are associated with such VPN routes, and are imported into different VRFs in other devices(see ). RD is allocated per VPN, each VRF imports one/multiple RTs(see ). The following sections will describe solutions to the above scenarios in detail.

In this scenario, PE1-PE4 and RR are iBGP peers. RD is allocated per VPN per PE. The offending VPN routes only carry one RT. We assume the network topology is shown in Figure 1.

When PE3 sends excessive VPN routes with RT1, while both PE1 and PE2 import VPN routes with RT1, the process of offending VPN routes will influence performance of VRFs on PEs. PEs and RR should have some mechanisms to identify and control the advertisement of offending VPN routes. a) PE1 If quota value is not set on PE1, and each VRF has a prefix limit on PE1. When the prefix limit of VPN1 VRF is exceeded, due to there is no other VRFs on PE1 need the VPN routes with RT1, PE1 will send VPN Prefix ORF message to RR and send warning to operator. The message will carry the prefix limit of VPN1 VRF, the RD value is set to 0 and the RT value is set to RT1. RR will withdraw the offending VPN routes and control the number of VPN routes sending to PE1. If quota value is set on PE1, each VRF has a prefix limit, and each <RD, source PE> tuple imported into VRF has a quota. When routes associated with <RD31, PE3> tuple pass the quota but the prefix limit of VPN1 VRF is not exceeded, PE1 sends a warning message to the operator, and the VPN Prefix ORF mechanism should not be triggered. After the prefix limit of VPN1 VRF is exceeded, since there is no other VRFs on PE1 need the VPN routes with RT1, PE1 will generate a BGP ROUTE-REFRESH message contains a VPN Prefix ORF entry, and send the entry to RR. RR will withdraw and stop to advertise such offending VPN routes (RD31, the prefix limit of VPN1 VRF, source PE is PE3, RT is RT1) to PE1. b) PE2 If quota value is not set on PE2, when the prefix limit of VPN1 VRF is exceeded, PE2 cannot trigger VPN Prefix ORF mechanism directly, because VPN2 VRF needs the VPN routes with RT1. PE2 triggers the mechanism only when the prefix limit for both, the VPN1 and VPN2 VRF has exceeded. The VPN Prefix ORF message will carry the VRF Prefix Limit = min(prefix limit of VPN1 VRF, prefix limit of VPN2 VRF), the RD value is set to 0 and the RT value is set to RT1. RR will withdraw the offending VPN routes and control the number of VPN routes sending to PE1. If quota value is set on PE2, both VPN1 VRF and VPN2 VRF import VPN routes with RT1. If PE2 triggers VPN Prefix ORF mechanism when VPN1 VRF overflows, VPN2 VRF cannot receive VPN routes with RT1 as well. PE2 should not trigger the VPN Prefix ORF mechanism to RR (RD31, min(prefix limit of VPN1 VRF, prefix limit of VPN2 VRF), RT1, RT2, source PE is PE3) until all the VRFs that import RT1 on it overflow.

In this scenario, PE1-PE4 and RR are iBGP peers. RD is allocated per VPN per PE. Multiple RTs are associated with the offending VPN routes, and are imported into different VRFs in other devices. We assume the network topology is shown in Figure 2.

When PE3 sends excessive VPN routes with RT1 and RT2, while both PE1 and PE2 import VPN routes with RT1, and PE1 also imports VPN routes with RT2. a) PE1 If quota value is not set on PE1, when the prefix limit of VPN1 VRF is exceeded, PE1 cannot trigger VPN Prefix ORF mechanism directly, because VPN2 VRF needs the VPN routes with RT1. PE1 triggers the mechanism only when the prefix limit for both, the VPN1 VRF and VPN2 VRF has exceeded. PE1 will send VPN Prefix ORF message to RR and send warning to operator. The message will carry the VRF Prefix Limit = min(prefix limit of VPN1 VRF, prefix limit of VPN2 VRF), the RD value is set to RD31, the RT value is set to RT1, RT2 and the source PE is PE3. RR will withdraw and stop advertising such offending VPN routes to PE1. If quota value is set on PE1, when routes associated with <RD31, PE3> tuple pass the quota but the prefix limit of VPN1 VRF is not exceeded, PE1 sends a warning to the operator. In case where the VPN1 VRF has exceeded its prefix limit and VPN2 VRF has not yet exceeded its prefix limit, PE1 should not send out the trigger message. Otherwise, the VPN2 VRF can't receive the VPN routes too (RR will filter all the VPN prefixes that contain RT1). PE1 should just discard the offending VPN routes locally. PE1 should only generate a BGP ROUTE-REFRESH message contains a VPN Prefix ORF entry(RD31, min(prefix limit of VPN1 VRF, prefix limit of VPN2 VRF), RT1, RT2, comes from PE3) in case where the prefix limit has exceeded for both the VRFs. b) PE2 If quota value is not set on PE2, when the prefix limit of VPN1 VRF is exceeded, due to there is no other VRFs on PE2 need the VPN routes with RT1, PE2 will send VPN Prefix ORF message to RR and send warning to operator. The message will carry the prefix limit of VPN1 VRF, the RD value is set to RD31 and the RT value is set to RT1. RR will withdraw and stop advertising such offending VPN routes to PE2. If quota value is set on PE2, due to there is only one VRF imports VPN routes with RT1. If it overflows, it will trigger VPN Prefix ORF (RD31, the prefix limit of VPN1 VRF, RT1, comes from PE3) mechanisms. RR will withdraw and stop advertising such offending VPN routes to PE2.

In this scenario, PE1-PE4 and RR are iBGP peers. RD is allocated per VPN. One/Multiple RTs are associated with the offending VPN routes, and be imported into different VRFs in other devices. We assume the network topology is shown in Figure 3.

When PE3 sends excessive VPN routes with RD1 and attached RT1 and RT2, while both PE1 and PE2 import VPN routes with RT1, the process of offending VPN routes will influence performance of VRFs on PEs. a) PE1 If quota value is not set on PE1, when the prefix limit of VPN1 VRF is exceeded, PE1 cannot trigger VPN Prefix ORF mechanism directly, because VPN2 VRF needs the VPN routes with RT2. PE1 triggers the mechanism only when the prefix limit for both, the VPN1 and VPN2 VRF has exceeded, and send warning to operator. The message will carry the VRF Prefix Limit = min(prefix limit of VPN1 VRF, prefix limit of VPN2 VRF), the RD value is set to RD1 and the RT value is set to RT1, RT2. RR will withdraw and stop advertising such offending VPN routes to PE1. If quota value is set on PE1, when routes associated with <RD1, PE3> tuple pass the quota but the prefix limit of VPN1 VRF is not exceeded, PE1 sends a warning to the operator. When the prefix limit of VPN1 VRF is exceeded, if the VPN2 VRF does not reach its limit at the same time, PE1 should still not send out the trigger message, because if it does so, the VPN2 VRF can't receive the VPN routes too (RR will filter all the VPN prefixes that contain RT1). PE1 should just discard the offending VPN routes locally. PE1 should only generate a BGP ROUTE-REFRESH message contains a VPN Prefix ORF entry(RD1, min(prefix limit of VPN1 VRF, prefix limit of VPN2 VRF), RT1, RT2, comes from PE3) when both the VRFs that import such prefixes are overflow. b) PE2 If quota value is not set on PE2, when the prefix limit of VPN1 VRF is exceeded, due to there is no other VRFs on PE2 need the VPN routes with RT1, PE2 will send VPN Prefix ORF message to RR and send warning to operator. The message will carry the prefix limit of VPN1 VRF, the RD value is set to RD1 and the RT value is set to RT1. RR will withdraw and stop advertising such offending VPN routes to PE2. If quota value is set on PE2, due to there is only one VRF imports VPN routes with RT1. If it overflows, it will trigger VPN Prefix ORF (RD1, the prefix limit of VPN1 VRF, RT1, comes from PE3) mechanisms. RR will withdraw and stop advertising such offending VPN routes to PE2. When PE2 overflows and PE1 does not overflow. PE2 triggers the VPN Prefix ORF message (RD1, RT1, comes from PE3). Using Source PE and RD, RR will only withdraw and stop advertising VPN routes with <RD1, RT1, come from PE3> to PE2. The communication between PE2 and PE1 for VPN1 will not be influenced.

We usually use NEXT_HOP to identify the source, but it may not useful in the following scenarios: a PE may have multiple addresses so that its BGP peer will receive several different NEXT_HOP from the same source. In Option B inter-domain scenario, the ASBR will change the NEXT_HOP. ORIGINATOR_ID is a non-transitive attribute generated by RR to identify the source, but ORIGINATOR_ID cannot be advertised outside the local AS. To cover the above scenarios, we define a new Extended Community: Source PE Extended Community(SPE EC) to transmit the identifier of source. Its value can be set by source PE/RR/ASBR. Once set and attached with the BGP UPDATE message, its value should not be changed along the advertisement path. The format of SPE EC is shown as Figure 4.

For the RR/ASBR, it should perform as following: Check the existence of the SPE EC. If it exists, does not change it. If SPE EC does not exist, check the existence of ORIGINATOR_ID. If it exists, put it into SPE EC. If ORIGINATOR_ID does not exist, put the router-id of source PE into SPE EC.

In this section, we defined a new ORF type called VPN Prefix Outbound Route Filter (VPN Prefix ORF). The ORF entries are carried in the BGP ROUTE-REFRESH message as defined in . A BGP ROUTE-REFRESH message can carry one or more ORF entries. The ROUTE-REFRESH message which carries ORF entries contains the following fields: AFI (2 octets) SAFI (1 octet) When-to-refresh (1 octet): the value is IMMEDIATE ORF Type (1 octet) Length of ORF entries (2 octets) A VPN Prefix ORF entry contains a common part and type-specific part. The common part is encoded as follows: Action (2 bits): the value is ADD, REMOVE or REMOVE-ALL Match (1 bit): the value is PERMIT or DENY Offending VPN routes process method (1 bit): if the value is set to 0, it means all offending VPN routes on the sender of VPN Prefix ORF message should be withdrawn; if the value is set to 1, it means the sender of VPN Prefix ORF message refuse to receive new offending VPN routes. The default value is 0. Reserved (4 bits) VPN Prefix ORF also contains type-specific part. The encoding of the type-specific part is shown in Figure 5.

Sequence: identifying the order in which RD-ORF is generated. Length: identifying the length of this VPN Prefix ORF entry. VRF Prefix Limit: carrying the prefix limt of the overflowed VRF. Route Distinguisher: distinguish the different user routes. The VPN Prefix ORF filters the VPN routes it tends to send based on Route Distinguisher. If RD is equal to 0, it means all VPN prefixes. Optional TLVs: carry the potential additional information to give the extensibility of the VPN Prefix ORF mechanism. Note that if the Action component of an ORF entry specifies REMOVE-ALL, the ORF entry does not include the type-specific part. When the BGP ROUTE-REFRESH message carries VPN Prefix ORF entries, it must be set as follows: The ORF-Type MUST be set to VPN Prefix ORF. The AFI MUST be set to IPv4, IPv6, or Layer 2 VPN (L2VPN). If the AFI is set to IPv4 or IPv6, the SAFI MUST be set to MPLS-labeled VPN address. If the AFI is set to L2VPN, the SAFI MUST be set to BGP EVPN. The Match field should be set to PERMIT when VRF Prefix Limit = 0xFFFF and RD=0; otherwise, the Match field should be set to DENY.

Source PE TLV is defined to identify the source of the VPN routes. For the sender of VPN Prefix ORF, it will check the existence of SPE EC. If it exists, the sender will put it into Source PE TLV. Otherwise, the value of Source PE TLV should be set to local AS number and NEXT_HOP. The source PE TLV contains the following types: Type = 1(suggested), Length = 8 octets, value = local AS number + NEXT_HOP. Type = 2(suggested), Length = 20 octets, value = local AS number + NEXT_HOP. Type = 3(suggested), Length = 8 octets, value = the value of AS number and ORIGINATOR_ID in Source PE Extended Community.

Route Target TLV is defined to identify the RT of the offending VPN routes. RT and RD can be used together to filter VPN routes when the source VRF contains multiple RTs, and the VPN routes with different RTs may be assigned to different VRFs on the receiver. The encoding of Route Target TLV is as follow: Type = 4(suggested), Length = 8*n (n is the number of RTs that the offending VPN routes attached) octets, value = the RT of the offending VPN routes. If multiple RTs are included, there must be an exact match.

The receiver of VPN Prefix ORF entries may be a RR or PE. As it receives the VPN Prefix ORF entries from the sender, it will check <AFI/SAFI, ORF-Type, Sequence, Route Distinguisher> to find if it already existed in its ORF-Policy table. If not, the receiver will add the VPN Prefix ORF entries into its ORF-Policy table; otherwise, the receiver will discard it. The RR or PE should determine the sending device and next hop address of the stored VPN Prefix ORF entry corresponding to the source PE in the filtering condition by looking up in the FIB, including: 1) determining the port field corresponding to the stored VPN Prefix ORF entry in the export routing filter strategy table; 2) Determine the sending device of the stored VPN Prefix ORF entry based on the value of the port field corresponding to the stored VPN Prefix ORF entry. The filtering conditions for the stored VPN Prefix ORF entries contain the RD and RT of the source PE. If the sending device of the VPN Prefix ORF entry stored under the same filtering condition includes all IBGP neighbors of the RR or PE other than the device with the next hop address, the VPN Prefix ORF entry is generated, where the next hop address is the next hop address sent to the network device in the direction of the source PE in the same filtering condition. The filtering condition in the generated VPN Prefix ORF entry contain the address, RD and RT of the source PE; Then, the RR or PE should send the generated VPN Prefix ORF entry to the device at the next hop address via BGP ROUTE-REFRESH message, so that the device at the next hop address filters the VPN route sent by the source PE. Before the route update is announced, the receiver should check its ORF-Policy table with <RD, Source PE> tuple of the VPN route. The Route Distinguisher information can be extracted directly from the BGP UPDATE message. The source PE information should be compared against the Source PE Extended Community if it is contained in BGP UPDATE message, or else the NEXT_HOP. If there is not a related VPN Prefix ORF entry in ORF-Policy table, the receiver will send this VPN route; otherwise, the receiver will perform the following operations: If the "Offending VPN routes process method" bit is 0, the receiver should withdraw all the VPN routes identified by RD, RT and information in optional TLVs in the entry, and stop sending the corresponding VPN routes to the sender. If the "Offending VPN routes process method" bit is 1, the receiver withdraw the extra VPN routes according to the value of VRF Prefix Limit, RD, RT and information in optional TLVs in the entry, and stop sending the corresponding VPN routes to the sender.

When the VPN Prefix ORF mechanism is triggered, the alarm information will be generated and sent to the network operators. Operators should manually configure the network to resume normal operation. Due to devices can record the VPN Prefix ORF entries sent by each VRF, operators can find the entries needs to be withdrawn, and trigger the withdraw process as described in manually. After returning to normal, the device sends withdraw ORF entries to its peer who have previously received ORF entries.

We take the scenario in as an example to illustrate how to determine each field when the sender generates a VPN Prefix ORF entry. We assume it is an IPv4 network. After PE1-PE4 and RR advertising the Outbound Route Filtering Capability, each of PE1-PE4 should send a VPN Prefix ORF entry that means "PERMIT-ALL" as follows: AFI is equal to IPv4 SAFI is equal to MPLS-labeled VPN address When-to-refresh is equal to IMMEDIATE ORF Type is equal to VPN Prefix ORF Length of ORF entries is equal to 26 Action is equal to ADD Match is equal to PERMIT Offending VPN routes process method is equal to 0 Sequence is equal to 0xFFFFFFFF Length is equal to 12 VRF Prefix Limit is equal to 0xFFFF Route Distinguisher is equal to 0 When the VPN Prefix ORF mechanism is triggered on PE1, PE1 will generate a VPN Prefix ORF contains the following information: AFI is equal to IPv4 SAFI is equal to MPLS-labeled VPN address When-to-refresh is equal to IMMEDIATE ORF Type is equal to VPN Prefix ORF Length of ORF entries is equal to 44 Action is equal to ADD Match is equal to DENY Offending VPN routes process method is equal to 0 Sequence is equal to 1 Length is equal to 30 VRF Prefix Limit is equal to the prefix limit of VPN1 VRF Route Distinguisher is equal to RD31 Optional TLV: Type is equal to 1 (Source PE TLV) Length is equal to 8 value is equal to PE3's IPv4 address Type is equal to 4 (Route Target TLV) Length is equal to 8 value is equal to RT1

This draft is experimental in order to determine if the proposed mechanism could block the offending routes as expected or not, and whether it would cause other potential network failures. The first section below describes implementation considerations for the mechanism. The second section below provides a short summary of the experimental topology and the results.

Before originating a VPN Prefix ORF message, the device should compare the list of RDs carried by VPN routes signaled for filtering and the RDs imported by not affected VRF(s). Once they have intersection, the VPN Prefix ORF message MUST NOT be originated. In deployment, the quota value can be set with different granularity, such as <PE>, <RD, Source AS>, etc. If the quota value is set to (VRF prefix limit/the number of PEs), whenever a new PE access to the network, the quota value should be changed. To avoid the frequently change of the quota value, the value can be set according to the following formula: Quota=MIN[(Margins coefficient)*<PE,CE limit>*<Number of PEs within the VPN, includes the possibility expansion in futures>, VRF Prefixes Limit] It should be noted that the above formula is only an example, the operators can use different formulas based on actual needs in management plane.

Currently, H3C has implemented VPN Prefix ORF mechanism related functions as follows: By configuring VRF Prefix limit and quota, achieve the use of RD and Source PE to control VPN routing. Generating, transmitting and processing Type 1 and Type 2 Source PE TLV. Using the Offending VPN routes process method to revoke all routes. Besides, we also implemented the following functions based on the open-source BGP implementation (FRR): VPN Prefix ORF mechanism triggered based on VRF limit in intra-domain and inter-domain scenarios. RD based VPN routing filtering in intra-domain and inter-domain scenarios.

The experiments will test whether the VPN Prefix ORF blocks the offending routes in the following scenarios: Intra-domain as a standalone mechanism, Inter-domain as a standalone mechanisms, Adding the VPN Prefix ORF to existing mechanisms for intra-domain VPNs, Adding the VPN Prefix ORF to existing mechanisms for intra-domain VPNs.

[TBD]

This draft does build upon . A BGP speaker will maintain the VPN Prefix ORF entries in an ORF-Policy table, this behavior consumes its memory and compute resources. To avoid the excessive consumption of resources, specifies that a BGP speaker can only accept ORF entries transmitted by its interested peers.

This document defines a new Outbound Route Filter type - VPN Prefix Outbound Route Filter (VPN Prefix ORF). The code point is from the "BGP Outbound Route Filtering (ORF) Types". It is recommended to set the code point of VPN Prefix ORF to 66. This document also define a VPN Prefix ORF TLV type under "Border Gateway Protocol (BGP) Parameters", four TLV types are defined:

This document also requests a new Transitive Extended Community Type. The new Transitive Extended Community Type name shall be "Source PE Extended Community".

Shunwan Zhuang Huawei Technologies Huawei Building, No.156 Beiqing Rd. Beijing Beijing, 100095 China

Thanks Robert Raszuk, Jim Uttaro, Jakob Heitz, Jeff Tantsura, Rajiv Asati, John E Drake, Gert Doering, Shuanglong Chen, Enke Chen, Srihari Sangli and Igor Malyushkin for their valuable comments on this draft.

The experimental topology is shown in Figure 6.

This topology can be used to verify as follows: whether the VPN Prefix ORF mechanism could block the offending routes in intra-domain scenario. whether the VPN Prefix ORF mechanism could block the offending routes in inter-domain scenario. whether the VPN Prefix ORF mechanism conflicts with the existing mechanism and cause failure. whether the quota value leads to flapping. TBD