| IDR | M. Marzetti |
| Internet-Draft | PCCW Global |
| Intended status: Standards Track | K. Patel |
| Expires: October 28, 2018 | Arrcus, Inc |
| J. Scudder | |
| Juniper Networks | |
| J. Heitz | |
| Cisco | |
| J. Snijders | |
| NTT | |
| April 26, 2018 |
Prefix Limit Based ORF for BGP-4
draft-keyur-idr-bgp-prefix-limit-orf-03
The BGP specification allows for "the ability to impose an (locally configured) upper bound on the number of address prefixes the speaker is willing to accept from a neighbor". In this specification, we define a new Outbound Route Filter type for BGP, termed "Prefix Limit Outbound Route Filter", which the speaker can use to communicate that upper bound to its peer. The peer is then required to abide by the limit. This is expected to have benefits in terms of resource consumption and more importantly, transparency of operation.
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 October 28, 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.
The Cooperative Outbound Route Filtering Capability defined in [RFC5291] provides a mechanism for a BGP speaker to send to its BGP neighbor a set of Outbound Route Filters (ORFs) that can be used by its neighbor to filter its outbound routing updates to the speaker.
This documents defines a new ORF-type for BGP, termed "Prefix Limit Outbound Route Filter (PrefixLimit ORF)", that can be used to perform Prefix Limit based route filtering. This filtering mechanism imposes a limit on a the number of unique prefixes that the BGP speaker can advertise to its neighbor.
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 RFC 2119.
The Prefix Limit ORF-Type allows a BGP speaker to inform its neighbor of its prefix limits. That is, it provides a mechanism through which a BGP speaker can request its neighbor to limit the number of unique prefixes that neighbor will advertise to the BGP speaker.
Conceptually a Prefix Limit ORF entry consists of the fields "Action, Match, Reserved, Prefix-Limit".
The value of the ORF-Type for the Prefix Limit ORF-Type is [TBD].
+--------------------------------------------------+
| Prefix-Limit (4 octets) |
+--------------------------------------------------+
A Prefix Limit ORF entry is encoded as follows. The "Action", "Match", and the "Reserved" field of the entry is encoded in the common part [RFC5291], and the remaining field of the entry is encoded in the "Type specific part" as follows.
A BGP speaker signals its compliance with this specification by listing the PrefixLimit ORF type in the Cooperative Route Filtering Capability as defined in [RFC5291].
We describe the rules for PrefixLimit primarily in terms of the rules for the router which sends a PrefixLimit ORF to its peer, which we term the "sending speaker", and for the router which receives a PrefixLimit ORF from its peer, which we term the "receiving speaker". Note that a given router may be either a sending or receiving speaker, or both, with respect to any given peering session.
A router which supports PrefixLimit ORF MUST keep track of the number of prefixes it has advertised to its peer -- when a new prefix is advertised, the count is incremented, and when a prefix is withdrawn, the count is decremented. A modification to the route for an already-advertised prefix does not change the count. We refer to this count as the "advertised prefix count" for the session. In effect, the advertised prefix count is equivalent to the size of the Adj-RIB-Out for the session.
A router which supports PrefixLimit ORF MAY maintain a received prefix count for its peer, which tracks the number of prefixes it has accepted from the peer. In effect, the received route count is equivalent to the size of the Adj-RIB-In for the session. The use of such a count is elaborated in the following section.
If a BGP speaker (the sending speaker) is configured to bound the number of prefixes it is willing to accept from its neighbor, it MAY advertise the value of that upper bound to that neighbor using PrefixLimit ORF. In this section and its subsection, when we refer to "the PrefixLimit" we are referring to the PrefixLimit value most recently advertised by the sending speaker to the receiving speaker.
If the sending speaker does not maintain a received prefix count, it is implicitly relying on its peer to correctly abide by this specification and no further action is required. If the sending speaker does maintain a received prefix count, it MAY locally enforce the PrefixLimit, according to the following rules.
When the sending speaker sends a PrefixLimit ORF which is less than its current received prefix count, it SHOULD wait for some interval before enforcing the new PrefixLimit. The interval to be used is a matter of local policy. Also, even if the PrefixLimit ORF is greater than or equal to the current received prefix count, the router may wish to wait for some interval before enforcing the new limit in order to allow for UPDATEs which may have been in flight prior to the receipt of the PrefixLimit ORF by the peer. Subsequent to any such waiting period, the remaining rules in this section SHALL apply.
If the PrefixLimit is exceeded (either because of a route announced by the peer or because the peer failed to timely withdraw routes after the PrefixLimit is revised downward), the peer is in violation, and the sending speaker MAY take corrective action. The router MAY also allow the received prefix count to exceed the PrefixLimit by some amount as a matter of local policy.
Corrective actions MAY include dropping the BGP session or refusing to accept new prefixes in excess of the PrefixLimit.
If the former option -- dropping the BGP session -- is chosen, the router MUST indicate this in advance by advertising its PrefixLimit ORF with the Match flag set to DENY. Also, by default it SHOULD NOT automatically reestablish the session.
If the latter option -- refusing to accept new prefixes -- is chosen, the router MUST accept modifications to already-accepted prefixes, and it MUST accept withdrawals of already-accepted prefixes. If prefixes are withdrawn, the received prefix count will drop below the announced PrefixLimit and new prefixes SHOULD be accepted, again up to but not exceeding the limit. Prefixes which are refused MUST NOT contribute to the received prefix count.
We note that the option of refusing to accept new prefixes will likely lead to desynchronization of the BGP session and is a flawed solution at best; operator intervention will be required in order to restore synchronization (for example, through correction of routing policies and a subsequent route-refresh).
When a PrefixLimit ORF is received, the new Prefix Limit value in the ORF is considered to be the new maximum Prefix Limit for the neighbor. In this section, when we refer to "the PrefixLimit" we are referring to the PrefixLimit value most recently received from the sending speaker by the receiving speaker.
If Match field is set to DENY the advertised Prefix-Limit value is purely informational and the receiving speaker SHOULD advertise the prefix even if doing so would cause its advertised prefix count to exceed the Prefix-Limit.
Else, if Match field is set to PERMIT, the receiving speaker MUST NOT advertise a prefix to its peer if doing so would cause its advertised prefix count to exceed the PrefixLimit.
The receiving speaker MAY take local action when its advertised prefix count approaches or reaches the PrefixLimit. The nature of the action (logging, incrementing counters, etc) is a matter of local policy, as is the threshold at which the action occurs.
When the receiving speaker receives a PrefixLimit ORF with When-to-Refresh set to DEFER, it need not take any additional action unless its current advertised prefix count exceeds the new PrefixLimit. In that case, it MUST take immediate steps to correct the violation.
Such steps MAY include withdrawing already-advertised prefixes so as to reduce the advertised prefix count to be less than or equal to the PrefixLimit. The selection of which prefixes to withdraw is a matter of local policy. Another option to correct the violation would be to drop the session; in this case the session SHOULD NOT be automatically reestablished.
When the receiving speaker receives a PrefixLimit ORF with When-to-Refresh set to IMMEDIATE, it behaves as given for DEFER but in addition advertises its Adj-RIB-Out as specified in [RFC5291].
ORFs provide information that guides future sending, but any malformed ORF is simply missed filtering information. If Prefix Limit ORF is malformed, then the Refresh messages shall simply be discarded.
This extension to BGP does not change the underlying security issues. However, it does suggest a mechanism by which certain denial of service risks may be reduced.
The authors would like to thank ... for their valuable comments.
IANA is requested to assign value TBD (PrefixLimit ORF) in the "BGP Outbound Route Filtering (ORF) Types" subregistry under the "Border Gateway Protocol (BGP) Parameters" registry.
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |
| [RFC5291] | Chen, E. and Y. Rekhter, "Outbound Route Filtering Capability for BGP-4", RFC 5291, DOI 10.17487/RFC5291, August 2008. |