RFC 8804 | CDNI Request Routing Extensions | September 2020 |
Finkelman & Mishra | Standards Track | [Page] |
Open Caching architecture is a use case of Content Delivery Network Interconnection (CDNI) in which the commercial Content Delivery Network (CDN) is the upstream CDN (uCDN) and the ISP caching layer serves as the downstream CDN (dCDN). This document defines extensions to the CDNI Metadata Interface (MI) and the Footprint & Capabilities Advertisement interface (FCI). These extensions are derived from requirements raised by Open Caching but are also applicable to CDNI use cases in general.¶
This is an Internet Standards Track document.¶
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at https://www.rfc-editor.org/info/rfc8804.¶
Copyright (c) 2020 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 Streaming Video Alliance [SVA] is a global association that works to solve streaming video challenges in an effort to improve end-user experience and adoption. The Open Caching Working Group [OCWG] of the Streaming Video Alliance [SVA] is focused on the delegation of video delivery requests from commercial CDNs to a caching layer at the ISP's network. Open Caching architecture is a specific use case of CDNI where the commercial CDN is the upstream CDN (uCDN) and the ISP caching layer is the downstream CDN (dCDN). The Open Caching Request Routing Functional Specification [OC-RR] defines the Request Routing process and the interfaces that are required for its provisioning. This document defines the CDNI metadata object [RFC8006] and the CDNI Footprint and Capabilities object [RFC8008] that are required for Open Caching Request Routing:¶
This document also registers CDNI Payload Types [RFC7736] for these defined objects.¶
For consistency with other CDNI documents, this document follows the CDNI convention of uCDN (upstream CDN) and dCDN (downstream CDN) to represent the commercial CDN and ISP caching layer, respectively.¶
The following terms are used throughout this document:¶
Additionally, this document reuses the terminology defined in [RFC6707], [RFC7336], [RFC8006], [RFC8007], and [RFC8008]. Specifically, we use the following CDNI acronyms:¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
Iterative CDNI Request Redirection is defined in Section 1.1 of [RFC7336] and elaborated by examples in Sections 3.2 and 3.4 of [RFC7336]. A Redirection Target (RT) is defined in Section 2 of [RFC7975] for Recursive Request Redirection as:¶
The endpoint to which the User Agent is redirected. In CDNI, an RT may point to a number of different components, some examples include a surrogate in the same CDN as the request router, a request router in a dCDN, or a surrogate in a dCDN.¶
In this document, we adopt the same definition of the RT for the Iterative Request Redirect use case. This use case requires the provisioning of the RT address to be used by the uCDN in order to redirect to the dCDN. RT addresses can vary between different footprints (for example, between different regions), and they may also change over time (for example, as a result of network problems). Given this variable and dynamic nature of the redirect target address, it may not be suitable to advertise it during bootstrap. A more dynamic and footprint-oriented interface is required. Section 4.3 of [RFC7336] suggests that it could be one of the roles of the FCI [RFC8008]. Following this suggestion, we have therefore chosen to use the CDNI Footprint & Capabilities Advertisement interface for redirect target address advertisement.¶
Use cases:¶
The Redirect Target capability object is used to indicate the target address the uCDN should use in order to redirect a client to the dCDN. A target may be attached to a specific uCDN host, attached to a list of uCDN hosts, or used globally for all the hosts of the uCDN.¶
When a dCDN is attaching the redirect target to a specific uCDN host or a list of uCDN hosts, the dCDN MUST advertise the hosts within the Redirect Target capability object as "redirecting-hosts". In this case, the uCDN can redirect to that dCDN address, only if the User Agent request was to one of these uCDN hosts.¶
If the Redirect Target capability object does not contain a target or the target is empty, the uCDN MUST interpret it as "no target available for these uCDN hosts for the specified footprint". In case such a target was already advertised in a previous FCI object, the uCDN MUST interpret it as an update that deletes the previous redirect target.¶
A redirect target for DNS redirection is an FQDN used as an alias in a CNAME record response (see [RFC1034]) of the uCDN DNS router. Note that DNS routers make routing decisions based on either the DNS resolver's IP address or the client IP subnet when EDNS0 client-subnet (ECS) is used (see [RFC7871]). The dCDN may choose to advertise redirect targets and footprints to cover both cases, such that the uCDN resolution would route the DNS query to different dCDN CNAMEs according to client subnet or dCDN resolver IP address. This method further allows the dCDN DNS to optimize the resolution by localizing the target CNAMEs. A uCDN implementation SHOULD prefer routing based on client IP subnet when the ECS option is present. A dCDN implementation using the ECS option MUST be aware of the privacy drawbacks listed in Section 2 of [RFC7871] and SHOULD follow the guidelines provided in Section 11.1 of [RFC7871].¶
A redirect target for HTTP redirection is the URI to be used as the value for the Location header of an HTTP redirect 3xx response, typically a 302 (Found) (see Section 7.1.2 of [RFC7231] and Section 6.4 of [RFC7231]).¶
The Redirect Target capability object consists of the following properties:¶
redirecting-hosts¶
dns-target¶
http-target¶
The following is an example of a Redirect Target capability object serialization that advertises a dCDN target address that is attached to a specific list of uCDN "redirecting-hosts". A uCDN host that is included in that list can redirect to the advertised dCDN redirect target. The capabilities object is serialized as a JSON object as defined in Section 5.1 of [RFC8008].¶
{ "capabilities": [ { "capability-type": "FCI.RedirectTarget", "capability-value": { "redirecting-hosts": [ "a.service123.ucdn.example.com", "b.service123.ucdn.example.com" ], "dns-target": { "host": "service123.ucdn.dcdn.example.com" }, "http-target": { "host": "us-east1.dcdn.example.com", "path-prefix": "/cache/1/", "include-redirecting-host": true } }, "footprints": [ <Footprint objects> ] } ] }¶
The DnsTarget object gives the target address for the DNS response to delegate from the uCDN to the dCDN.¶
host¶
The following is an example of the DnsTarget object:¶
{ "host": "service123.ucdn.dcdn.example.com" }¶
The following is an example of a DNS query for uCDN address "a.service123.ucdn.example.com" and the corresponding CNAME redirection response:¶
Query: a.service123.ucdn.example.com: type A, class IN Response: NAME: a.service123.ucdn.example.com, TYPE: CNAME, CLASS: IN, TTL: 120, RDATA: service123.ucdn.dcdn.example.com¶
The HttpTarget object gives the necessary information to construct the target Location URI for HTTP redirection.¶
host¶
scheme¶
path-prefix¶
include-redirecting-host¶
The following is an example of the HttpTarget object with a "scheme", a "path-prefix", and "include-redirecting-host" properties:¶
{ "host": "us-east1.dcdn.example.com", "scheme": "https", "path-prefix": "/cache/1/", "include-redirecting-host": true }¶
The following is an example of an HTTP request for content at uCDN host "a.service123.ucdn.example.com" and the corresponding HTTP response with a Location header, used for redirecting the client to the dCDN, constructed according to the HttpTarget object from the above example:¶
Request: GET /vod/1/movie.mp4 HTTP/1.1 Host: a.service123.ucdn.example.com Response: HTTP/1.1 302 Found Location: https://us-east1.dcdn.example.com/cache/1/ a.service123.ucdn.example.com/vod/1/movie.mp4¶
Before requests can be routed from the uCDN to the dCDN, the CDNs must exchange service configurations between them. Using the MI, the uCDN advertises out-of-band its hosts to the dCDN; each host is designated by a hostname and has its own specific metadata (see Section 4.1.2 of [RFC8006]). Using the FCI, the dCDN advertises (also out-of-band) the redirect target address defined in Section 2.3 for the relevant uCDN hosts. The following is a generalized example of the message flow between a uCDN and a dCDN. For simplicity, we focus on the sequence of messages between the uCDN and dCDN and not on how they are passed.¶
Explanation:¶
Once the redirect target has been set, the uCDN can start redirecting user requests to the dCDN. The following is a generic sequence of redirection using the host and redirect target that were advertised in Figure 1.¶
Explanation:¶
Open Caching requires that the uCDN provides a fallback target server to the dCDN to be used in cases where the dCDN cannot properly handle the request. To avoid redirect loops, the fallback target server's address at the uCDN MUST be different from the original uCDN address from which the client was redirected to the dCDN. The uCDN MUST avoid further redirection when receiving the client request at the fallback target. The Fallback Target is defined as a generic metadata object (see Section 3.2 of [RFC8006]).¶
Use cases:¶
The Fallback Target metadata object is used to indicate the target address the dCDN should redirect a client to when falling back to the uCDN. The fallback target address is represented as an Endpoint object as defined in Section 4.3.3 of [RFC8006].¶
In DNS redirection, a CNAME record is used as the fallback target address.¶
In HTTP redirection, a hostname is used as the fallback target address.¶
When using HTTP redirect to route a client request back to the uCDN, it is the dCDN's responsibility to use the original URL path as the client would have used for the original uCDN request, stripping, if needed, the dCDN path-prefix and/or the uCDN hostname from the redirect URL that may have been used to request the content from the dCDN.¶
The MI.FallbackTarget generic metadata object consists of the following two properties:¶
host¶
scheme¶
The following is an example of an MI.FallbackTarget generic metadata object that designates the host address the dCDN should use as fallback address to redirect back to the uCDN:¶
{ "generic-metadata-type": "MI.FallbackTarget", "generic-metadata-value": { "host": "fallback-a.service123.ucdn.example", "scheme": "https" } }¶
The uCDN advertises out-of-band the fallback target address to the dCDN, so that the dCDN may redirect a request back to the uCDN in case the dCDN cannot serve it. Using the MI, the uCDN advertises its hosts to the dCDN, along with their specific host metadata (see Section 4.1.2 of [RFC8006]). The Fallback Target generic metadata object is encapsulated within the "host-metadata" property of each host. The following is an example of a message flow between a uCDN and a dCDN. For simplicity, we focus on the sequence of messages between the uCDN and dCDN, not on how they are passed.¶
Explanation:¶
The following is a generic sequence of redirection using the configurations that were advertised in Figure 3. In this case, the dCDN redirects back to the uCDN fallback target address.¶
Explanation:¶
When advertising fallback addresses to the dCDN, the uCDN SHOULD consider the failure use cases that may lead the dCDN to route requests to uCDN fallback. In extreme dCDN network failures or under denial-of-service (DoS) attacks, requests coming from a large segment or multiple segments of the dCDN may be routed back to the uCDN. The uCDN SHOULD therefore design its fallback addressing scheme and its available resources accordingly. A favorable approach would be for the uCDN to use a different fallback target address for each uCDN host, enabling it to load balance the requests using the same methods as it would for its original hosts. See Sections 4.1.2 and 4.1.3 of [RFC8006] for a detailed description of how to use GenericMetadata objects within the HostMatch object advertised in the HostIndex of the uCDN.¶
IANA has registered the following CDNI Payload Types in the "CDNI Payload Types" registry defined in [RFC7736]:¶
Payload Type | Specification |
---|---|
FCI.RedirectTarget | RFC 8804 |
MI.FallbackTarget | RFC 8804 |
This specification defines extensions to the CDNI Metadata Interface (MI) and the Footprint & Capabilities Advertisement interface (FCI). As such, it is subject to the security and privacy considerations defined in Section 8 of [RFC8006] and in Section 7 of [RFC8008], respectively.¶
The Redirect Target capability object potentially reveals information about the internal structure of the dCDN network. A third party could intercept the FCI transactions and use the information to attack the dCDN. The same is also true for the Fallback Target generic metadata object, as it may reveal information about the internal structure of the uCDN, exposing it to external exploits. Implementations of the FCI and MI MUST therefore use strong authentication and encryption and strictly follow the directions for securing the interface as defined for the Metadata Interface in Section 8.3 of [RFC8006].¶
The authors thank Nir B. Sopher for reality checks against production use cases; his contribution is significant to this document. The authors also thank Ben Niven-Jenkins for his review and feedback and Kevin J. Ma for his guidance throughout the development of this document, including his regular reviews.¶