| PCE Working Group | Q. Wu |
| Internet-Draft | D. Dhody |
| Intended status: Standards Track | Huawei |
| Expires: October 31, 2015 | D. King |
| Lancaster University | |
| D. Lopez | |
| Telefonica I+D | |
| J. Tantsura | |
| Ericsson | |
| April 29, 2015 |
Path Computation Element (PCE) Discovery using Domain Name System(DNS)
draft-wu-pce-dns-pce-discovery-08
Discovery of the Path Computation Element (PCE) within an IGP area or routing domain is possible using OSPF and IS-IS IGP discovery. However, it has been established that in certain deployment scenarios PCEs may not wish, or be able to participate within the IGP process. In those scenarios, it is beneficial for the Path Computation Client (PCC) (or other PCE) to discover PCEs via an alternative mechanism to using an IGP discovery.
This document specifies the requirements, use cases, procedures and extensions to support PCE discovery along with certain relevant information type and capability discovery via DNS.
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 October 31, 2015.
Copyright (c) 2015 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.
The Path Computation Element Communication Protocol (PCEP) is a transaction-based protocol carried over TCP [RFC4655]. In order to be able to direct path computation requests to the Path Computation Element (PCE), a Path Computation Client (PCC) (or other PCE) needs to know the location and capability of a PCE.
In a network where an IGP is used and where the PCE participates in the IGP, discovery mechanisms exist for PCC (or PCE) to learn the identity and capability of each PCE. [RFC5088] defines a PCE Discovery (PCED) TLV carried in an OSPF Router LSA. Similarly, [RFC5089] defines the PCED sub-TLV for use in PCE Discovery using IS-IS. Scope of the advertisement is limited to IGP area/level or Autonomous System (AS).
However in certain scenarios not all PCEs will participate in the same IGP instance, section 3 (Motivation) outlines a number of use cases. In these cases, current PCE Discovery mechanisms are therefore not appropriate and another PCE discovery function would be required. (sec 4 of [PCE-QUESTION]).
This document describes PCE discovery via DNS. The mechanism with which DNS comes to know about the PCE and its capability is out of scope of this document.
The following terminology is used in this document.
As described in [RFC4674], the PCE Discovery information should at least be composed of:
These PCE discovery information allows PCCs to select appropriate PCEs.
This document specifies the procedures and extension to facilitate DNS-based PCE information discovery for specific use cases, and to complement existing IGP discovery mechanism.
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 [RFC2119].
This section discusses in more detail the motivation and use cases for an alternative DNS-based PCE discovery mechanism.
When the PCE is a router participating in the IGP, or even a server participating passively in the IGP, with all PCEP speakers in the same routing domain, a simple and efficient way to announce PCEs consists of using IGP flooding.
It has been identified that the existing PCE discovery mechanisms do not work very well in following scenarios:
Advertisement based PCE discovery using IGP methods [RFC5088] and [RFC5089] floods the PCE information to an area, a subset of areas or to a full routing domain. By the very nature of flooding and advertisements it generates unwanted traffic and may lead to unnecessary advertisement, especially when PCE information needs frequent changes.
DNS is a query-response based mechanism, a client (a PCC) can use DNS to discover a PCE only when it needs to compute a path and does not require any other node in the network to be involved.
In case of Intermittent PCEP session, where PCEP sessions are systematically open and closed for each PCEP request, a DNS-based query-response mechanism is more suitable. One may also utilize DNS-based load-balancing and recovery functions.
Server virtualization has gain importance since it provides better reliability and high availability in the event of hardware failure. It allows for higher utilization of physical resources while improving administration by having a single management interface for all virtual servers.
When one PCE instance is virtually hosted on a server and initiated as a PCE instance, another PCE instance may be created on the same server or a different server to provide better load balancing and reliability. In such a case, where there are a large number of PCCs that need to know these PCE instances' location, manual configuration on PCCs for PCC and PCE relationship is not trivial or desirable.
In the case of H-PCE ,when a dynamic Address is assigned to the parent PCE, any existing configuration entry on child PCE becomes invalid and the parent PCE becomes unreachable. In order to handle changes in parent PCE identities, the DNS update can be used to provide IP reachability to the parent PCE with new assigned Address. The DNS update can be performed by either parent PCE or OSS/NMS that is aware of PCE Identities changes.
Applications make use of DNS lookups on FQDN to find a node(e.g., PCEP endpoint). When a PCE performs DNS lookup or dynamic DNS update with the DNS server, the PCE MUST have a security association of some type with the DNS server. The security association SHOULD be established either using DNSSEC [RFC4033] or TSIG/TKEY[RFC2845][RFC2930]. DNS lookup for PCE Discovery can be applied either within an administration domain or spanning across administration domains. A security association is REQUIRED even if the DNS server is in the same administrative domain as the PCE.
Multiple PCEs can be present in a single network domain for redundancy. DNS supports inherent load balancing where multiple PCEs (with different IP addresses) are known in DNS for a single PCE server name and are hidden from the PCC.
In an IGP advertisement based PCE discovery, one learns of all the PCEs and it is the job of the PCC to do load-balancing.
A DNS-based load-balancing mechanism works well in case of Intermittent PCEP sessions and request are load-balanced among PCEs similar to HTTP request without any complexity at the client.
service-parms = [ [app-service] *(":" app-protocol)]
app-service = experimental-service / iana-registered-service
app-protocol = experimental-protocol / iana-registered-protocol
experimental-service = "x-" 1*30ALPHANUMSYM
experimental-protocol = "x-" 1*30ALPHANUMSYM
iana-registered-service = ALPHA *31ALPHANUMSYM
iana-registered-protocol = ALPHA *31ALPHANUMSYM
ALPHA = %x41-5A / %x61-7A ; A-Z / a-z
DIGIT = %x30-39 ; 0-9
SYM = %x2B / %x2D / %x2E ; "+" / "-" / "."
ALPHANUMSYM = ALPHA / DIGIT / SYM
; The app-service and app-protocol tags are limited to 32
; characters and must start with an alphabetic character.
; The service-parms are considered case-insensitive.
The NAPTR service field format defined by the S-NAPTR DDDS application in [RFC3958] follows this Augmented Backus-Naur Form (ABNF) [RFC5234]:
iana-registered-service =/ pce-service
pce-service = "pce" *("+" appln-name)
appln-name = non-ws-string
non-ws-string = 1*(%x21-FF)
This specification refines the "iana-registered-service" tag definition for the discovery of PCE supporting a specific PCE application or multiple PCE applications as defined below.
The appln-name element is the Application Identifier used to identify a specific PCE application. The PCE Application Name are allocated by IANA as defined in section 8.1.
iana-registered-protocol =/ pce-protocol
pce-protocol = "pce." pce-transport
pce-transport = "tcp" / "tls.tcp"
This specification also refines the "iana-registered-protocol" tag definition for the discovery of PCE supporting a specific transport protocol as defined below.
Similar to application protocol tags defined in the [RFC6408],the S-NAPTR application protocol tags defined by this specification MUST NOT be parsed in any way by the querying application or Resolver. The delimiter (".") is present in the tag to improve readability and does not imply a structure or namespace of any kind. The choice of delimiter (".") for the application protocol tag follows the format of existing S-NAPTR application protocol tag registry entries, but this does not imply that it shares semantics with any other specifications that create registry entries with the same format.
The S-NAPTR application service and application protocol tags defined by this specification are unrelated to the IANA "Service Name and Transport Protocol Port Number Registry" (see [RFC6335]).
The maximum length of the NAPTR service field is 256 octets, including a one-octet length field (see Section 4.1 of [RFC3403] and Section 3.3 of [RFC1035]).
A PCE Client MUST be capable of using the extended S-NAPTR application service tag for dynamic discovery of a PCE supporting Standards Track applications. Therefore, every IETF Standards Track PCE application MUST be associated with a "PCE-service" tag formatted as defined in this specification and allocated in accordance with IANA policy (see Section 8).
'PCE+gco:pce.tcp'
For example, a NAPTR service field value of:
means that the PCE in the SRV or A/AAAA record supports the Global Concurrent Optimization Application (See section 8.1)and the Transport Control Protocol (TCP) as the transport protocol (See section 8.2).
Domain Name System (DNS) administrators SHOULD also provision legacy NAPTR records [RFC3403] in order to guarantee backwards compatibility with legacy PCE that only support S-NAPTR DDDS application in [RFC3958]. If the DNS administrator provisions both extended S-NAPTR records as defined in this specification and legacy NAPTR records defined in [RFC3403], then the extended S-NAPTR records MUST have higher priority(e.g., lower order and/or preference values) than legacy NAPTR records.
The extended-format NAPTR records provide a mapping from a domain to the SRV record or A/AAAA record for contacting a server supporting a specific transport protocol and PCE application. The resource record will contain an empty regular expression and a replacement value, which is the SRV record or the A/AAAA record for that particular transport protocol.
The assumption for this mechanism to work is that the DNS administrator of the queried domain has first provisioned the DNS with extended-format NAPTR entries.
When the PCC or other PCEs performs a NAPTR query for a server in a particular realm, the PCC or other PCEs has to know in advance the search path of the resolver, i.e.,in which realm to look for a PCE, and in which Application Identifier it is interested.
The search path of the resolver can either be pre-configured, or discovered using Diameter, DHCP or other means. For example, the realm could be deduced from the Network Access Identifier (NAI) in the User-Name attribute-value pair (AVP) or extracted from the Destination-Realm AVP in Diameter [RFC6733].
When pre-configuration is used, PCE domain(e.g.,AS200)can be added as "subdomains" of the first-level domain of the underlying service (e.g., AS200.example.com), which allows a NAPTR query for a server in a PCE domain associated with DNS domain-name.
When DHCP is used, it SHOULD know the domain-name of that realm and use DHCP to discover IP address of the PCE in that realm that provides path computation service along with some PCE location information useful to a PCC (or other PCE) for a PCE selection, and contact it directly. In some instances, the discovery may result in a per protocol/application list of domain-names that are then used as starting points for the subsequent S-NAPTR lookups [RFC3958]. If neither the IP address nor other PCE location information can be discovered with the above procedure, the PCC (or other PCE) MAY request a domain search list, as described in [RFC3397] and[RFC3646], and use it as input to the DDDS application.
When the PCC (or other PCE) does not find valid domain-names using the mechanisms above, it MUST stop the attempt to discover any PCE.
The following procedures result in an IP address, PCE domain, neighboring PCE domain and PCE Computation Scope where the PCC (or other PCE) can contact the PCE that hosts the service it is looking for.
The PCC (or other PCE) should know the service identifier for the Path Computation service and associated transport protocol. The service identifier for the Path Computation service is defined as "PCE+apX" as specified in section 5, The PCE supporting "PCE" service MUST support TCP as transport, as described in [RFC5440].
The services relevant for the task of transport protocol selection are those with S-NAPTR service fields with values "PCE+apX:Y", where 'PCE+apX' is the service identifier defined in the previous paragraph, and ' Y' is the letter that corresponds to a transport protocol supported by the PCE. This document also establishes an IANA registry for mappings of S-NAPTR service name to transport protocol.
These NAPTR [RFC3958] records provide a mapping from a domain to the SRV [RFC2782] record for contacting a PCE with the specific transport protocol in the S-NAPTR services field. The resource record MUST contain an empty regular expression and a replacement value, which indicates the domain name where the SRV record for that particular transport protocol can be found. As per [RFC3403], the client discards any records whose services fields are not applicable.
The PCC (or other PCE) MUST discard any service fields that identify a resolution service whose value is not valid. The S-NAPTR processing as described in [RFC3403] will result in the discovery of the most preferred PCE that is supported by the client, as well as an SRV record for the PCE.
If the returned NAPTR service fields contain entries formatted as "pce+apX:Y" where "X" indicates the Application Identifier and "Y" indicates the supported transport protocol(s), the target realm supports the extended format for NAPTR-based PCE discovery defined in this document.
If the returned NAPTR service fields contain entries formatted as "pce+apX" where "X" indicates the Application Identifier, the target realm supports the extended format for NAPTR-based PCE discovery defined in this document.
If the returned NAPTR service fields contain entries formatted as "pce:X" where "X" indicates the supported transport protocol(s), the target realm supports PCEP but does not support the extended format for NAPTR-based PCE discovery defined in this document.
If the returned NAPTR service fields contain entries formatted as "pce", the target realm supports PCEP but does not support the extended format for NAPTR-based PCE discovery defined in this document. The PCEP implementation resolves the "replacement" field entry to a target host using the lookup method appropriate for the "flags" field and attempts to connect using TCP (in future it SHOULD attempt all supported transport Protocols) .
Note that the regexp field in the S-NAPTR example above is empty. The regexp field MUST NOT be used when discovering PCE, as its usage can be complex and error prone. Also, the discovery of the PCE does not require the flexibility provided by this field over a static target present in the TARGET field.
As the default behavior, the client is configured with the information about which transport protocol is used for a path computation service in a particular domain. The client can directly perform an SRV query for that specific transport using the service identifier of the path computation Service. For example, if the client knows that it should be using TCP for path computation service, it can perform a SRV query directly for_PCE._tcp.example.com.
Once the server providing the desired service and the transport protocol has been determined, the next step is to determine the IP address.
According to the specification of SRV RRs in [RFC2782], the TARGET field is a fully qualified domain-name (FQDN) that MUST have one or more address records; the FQDN must not be an alias, i.e., there MUST NOT be a CNAME or DNAME RR at this name. Unless the SRV DNS query already has reported a sufficient number of these address records in the Additional Data section of the DNS response (as recommended by [RFC2782]), the PCC needs to perform A and/or AAAA record lookup(s) of the domain-name, as appropriate. The result will be a list of IP addresses, each of which can be contacted using the transport protocol determined previously.
Order Pref Flags Service Regexp Replacement
IN NAPTR 50 50 "s" "pce:pce.tls.tcp" ""
_PCE._tcp.as100.example.com
IN NAPTR 90 50 "s" "pce:pce.tcp" ""
_PCE._tcp.as100.example.com
As an example, consider a client that wishes to find PCED service in the as100.example.com domain. The client performs a S-NAPTR query for that domain, and the following NAPTR records are returned:
;; Priority Weight Port Target
IN SRV 0 1 XXXX server1.as100.example.com
IN SRV 0 2 XXXX server2.as100.example.com
This indicates that the domain does have a PCE providing Path Computation services over TCP, in that order of preference. If the client only supports TCP, TCP will be used, targeted to a host determined by an SRV lookup of _PCE._tcp.example.com. That lookup would return:
;; order pref flags service regexp replacement
IN NAPTR 150 50 "a" "pce:pce.tcp" ""
server1.ex2.example.com
IN NAPTR 150 50 "a" "pce:pce.tls.tcp" ""
server2.ex2.example.com
IN NAPTR 150 50 "a" "pce+gco:pce.tcp" ""
server1.ex2.example.com
IN NAPTR 150 50 "a" "pce+gco:pce.tls.tcp" ""
server2.ex2.example.com
As an alternative example, a client wishes to discover a PCE in the ex2.example.com realm that supports the GCO application over TCP. The client performs a NAPTR query for that domain, and the following NAPTR records are returned:
This indicates that the server supports GCO(ID=1) over TCP and TLS/TCP via hosts server1.ex2.example.com and server2.ex2.example.com, respectively.
DNS servers MAY use DNS TXT record to give additional information about PCE service and add such TXT record to the additional information section (See section 4.1 of [RFC1035]) that are relevant to the answer and have the same authenticity as the data (Generally this will be made up of A and SRV records)in the answer section. The additional information may include path computation capability, the PCE domains and Neighbor PCE domains associated with the PCE. If discovery of PCE supporting a specific PCE capability described in section 7.2 has already been performed, capability associated with the PCE does not need to be included in the additional information.
<owner> <class> <ttl> TXT "<attribute name>=<attribute value>"
To store new types of information, the TXT record uses a structured format in its TXT-DATA field [RFC1035]. The format consists of the attribute name followed by the value of the attribute. The name and value are separated by an equals sign (=). The general syntax may follow one defined in section 2 of [RFC1464] as follows:
ex2.example.com IN TXT "pce domain = as10" ex2.example.com IN TXT "neigh domain= as5" ex2.example.com IN TXT "cap=link constraint"
For example, the following TXT records contain attributes specified in this fashion:
The client MAY inspect those Additional Information section in the DNS message and be capable of handling responses from nameservers that never fill in the Additional Information part of a response.
+------------------+----------------------------+ | Tag | PCE Application | +------------------+----------------------------+ | pce+gco | GCO [RFC5557] | | pce+p2mp | P2MP [RFC5671] | | pce+stateful | Stateful [STATEFUL-PCE] | | pce+gmpls | GMPLS [RFC7025] | | pce+interas | Inter-AS[RFC5376] | | pce+interarea | Inter-Area [RFC4927] | | pce+interlayer | Inter-layer [RFC6457] | +------------------+----------------------------+
IANA specifies to create a new registry ‘ S-NAPTR application service tags’ for existing IETF PCE applications.
Future IETF PCE applications MUST reserve the S-NAPTR application service tag corresponding to the allocated PCE Application ID as defined in Section 3.
+------------------+----------+
| Tag | Protocol |
+------------------+----------+
| pce.tcp | TCP |
+------------------+----------+
IANA has reserved the following S-NAPTR Application Protocol Tags for the PCE transport protocols in the "S-NAPTR Application Protocol Tag" registry created by [RFC3958].
Future PCE versions that introduce new transport protocols MUST reserve an appropriate S-NAPTR Application Protocol Tag in the "S-NAPTR Application Protocol Tag" registry created by [RFC3958].
This document specifies an enhancement to the NAPTR service field format. The enhancement and modifications are based on the S-NAPTR, which is actually a simplification of the NAPTR, and therefore the same security considerations described in [RFC3958] are applicable to this document.
For most of those identified threats, the DNS Security Extensions [RFC4033] does provide protection. It is therefore recommended to consider the usage of DNSSEC [RFC4033] and the aspects of DNSSEC Operational Practices [RFC6781] when deploying Path Computation Services.
In deployments where DNSSEC usage is not feasible, measures should be taken to protect against forged DNS responses and cache poisoning as much as possible. Efforts in this direction are documented in [RFC5452].
However a malicious host doing S-NAPTR queries learns applications supported by PCEs in a certain realm faster, which might help the malicious host to scan potential targets for an attack more efficiently when some applications have known vulnerabilities.
Where inputs to the procedure described in this document are fed via DHCP, DHCP vulnerabilities can also cause issues. For instance, the inability to authenticate DHCP discovery results may lead to the Path Computation service results also being incorrect, even if the DNS process was secured.
The author would like to thank Claire Bi,Ning Kong, Liang Xia, Stephane Bortzmeyer,Yi Yang, Ted Lemon, Adrian Farrel and Stuart Cheshire for their review and comments that help improvement to this document.