| Internet-Draft | SRv6 Addressing | October 2025 |
| Horn, et al. | Expires 23 April 2026 | [Page] |
This document describes SRv6 addressing for NEXT-CSID locator format. It introduces concepts of Blocks, Sets, and Node IDs, and explains how summarization boundaries and flexible algorithm support can be implemented for both small and large networks.¶
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 23 April 2026.¶
Copyright (c) 2025 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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Segment Routing over IPv6 (SRv6) enables the creation of source routing policies by encoding instructions in the form of IPv6 addresses, known as Segment Identifiers (SIDs). To efficiently manage SRv6 deployments at scale, it is critical to design a structured and summarizable addressing plan. This document focuses on SRv6 deployments using the NEXT-CSID compression format (RFC9800) which is the most efficient and widely deployed SRv6 compression. Addressing structure is based on an SRv6 SID structure with 32-bit Locator-Block length and 16-bit CSID length, or F3216 locator format for short, and provides operational guidelines for Service Providers network of different sizes.¶
This document leverages the terms defined in [RFC8402], [RFC8754], [RFC8986], [RFC9350], and [RFC9800], in particular segment, segment list, Segment Identifier (SID), SID list, SR policy, prefix segment, adjacency segment, SRH, SR domain, SR source node, SR segment endpoint node, transit node, SRv6 endpoint behavior, flavor, SID block, locator, function, and argument, IGP Flexible Algorithm (FA), Locator-Block, Locator-Node, Compressed-SID (CSID), CSID container, CSID length, compressed SID list, Global Identifier Block (GIB), and Locator Identifier Block (LIB). The reader is assumed to be familiar with this terminology.¶
This document introduces the following new terms:¶
SRv6 locator is a routable prefix allocated to a node and associated with a single algorithm. An SRv6 locator is often called simply "locator".¶
Infrastructure addresses and SRv6 locators serve different purposes and perform different roles in the network. Therefore, SRv6 locators must be allocated from a separate address range to facilitate network operation and security.¶
Since SRv6 locators have different allocation requirements and constraints, they should be allocated using a dedicated SRv6 addressing plan. An existing IPv6 addressing plan should not be a constraint for the SRv6 addressing plan.¶
This document uses the SRv6 SID address block 5f00::/16 ([RFC9602]), but the same principles are applicable when using ULA or GUA address ranges.¶
This document focuses on the F3216 locator format, currently the most widely deployed format due to its compression efficiency and simplicity. But all recommendations are applicable to any other format.¶
It is valuable to use nibble boundaries whenever possible to improve human readability.¶
[RFC9800] structures an SRv6 locator as the combination Locator-Block:Locator-Node. For simplicity, we will use the terms Block ID and CSID to refer to Locator-Block and Locator-Node respectively.¶
Format F3216 uses 32 bits for the Block ID and 16 bits for the CSID. This results in a 16-bit Locator-Node length.¶
This locator structure can be represented as follows, using one character for each 4 bits:¶
BBBB:BBBB:NNNN::/48¶
Where:¶
[RFC9800] specifies the Global ID Block (GIB) and Local ID Block (LIB) as two non-overlapping sub-spaces of the CSID numbering space.¶
"The CSID values that are allocated from the GIB have a global semantic within the Locator-Block, while those that are allocated from the LIB have a local semantic on an SR segment endpoint node and within the scope of the Locator-Block."¶
[RFC9800] states that, for a CSID allocated from the GIB:¶
"The tuple (Locator-Block, CSID) identifies the same segment across all nodes of the SR domain."¶
And, for a CSID allocated from the LIB:¶
"The tuple (Locator-Block, CSID) identifies a different segment on each node of the SR domain."¶
CSID values allocated from the GIB represent globally unique segments and can be included as part of the locator. On the other hand, CSID values allocated from the LIB represent segments that are locally significant to each node and, as a such, cannot be used as part of a locator. This separation is fundamental to ensuring fully deterministic behavior for TE.¶
The boundary between the GIB and LIB is flexible, but it must be consistent across all nodes of the SR domain.¶
The industry-accepted default sub-spaces for the F3216 format uses the uSID values 0x0001 to 0xdfff for GIB and the SIDs 0xe000 to 0xffff for LIB. This implies that for each Block there are 57,344 global SIDs (excluding CSID value 0, which is reserved to mark the end-of-container) and 8,192 local SIDs.¶
To organize address spaces more efficiently, the CSID numbering space within a block is divided into smaller, equally sized sections called sets. Each set has a unique Set ID. The remaining less significant bits of the CSID space are used for Node IDs, to uniquely identify each node within a set. The combination of the Block ID and Set ID creates a subnet prefix. This subnet can be assigned to a specific routing area or domain. This approach makes logical address assignment and address summarization much simpler and more efficient.¶
A Set ID can be any length, but for human readability it is recommended to use sizes that align with nibble (4-bit) boundaries. The most common size for a Set ID is 8 bits.¶
The locator structure is as follows:¶
BBBB:BBBB:SSNN::¶
Where:¶
This structure supports 256 different sets within each block. Each set can contain up to 256 unique Node IDs.¶
When using the default split between GIB and LIB, then Set IDs from 0x00 to 0xdf are used for locators (globally significant) and Set IDs from 0xe0 to 0xff are used for local functions.¶
An area is any part of the network with defined boundaries where address summarization can be performed (such as an ISIS level or an OSPF area). The number of devices in an area can vary, typically containing anywhere from a few hundred to several thousand devices.¶
Operators assign an appropriate number of sets to each area, based on the number of devices it needs to support.¶
Set Assignment Example:¶
Block ID: 5f00:0000::/32¶
Area 1 - 500 devices:¶
Area 2 - 1500 devices:¶
Sets assigned to single area do not have to be consecutive. (See the Summarization section for more details.)¶
Summarizing locators is essential for achieving virtually unlimited scalability in an SRv6 network. The need for summarization depends on the size of the network and the scalability of individual devices. In very small networks, summarization may not be necessary.¶
It proves to be operationally very beneficial to use a unified summarization scheme across the entire network. No matter the area size, always summarize at set boundaries. This approach keeps routing tables simple, clean, and structured, while still providing strong summarization benefits. It also eases the preservation of addressing space for future growth.¶
Summarization Example:¶
Area 1 - summaries:¶
Area 2 - summaries:¶
It is technically possible to advertise larger blocks, such as 5f00:0000:0000::/39 for Area 1 or 5f00:0000:0200::/38 for Area 2. However, the summarization gain here is negligible, while increasing operational complexity.¶
A small network is any network where every device can be assigned a locator from a single block. Theoretically, a single block can support up to 57,344 devices. In practice it is suitable for networks with up to around 35,000 devices.¶
In this scenario, only one block is used to address all devices in the network. This is the optimal option for TE, because all SIDs in any policy will come from the same block. As a result, a headend can use the minimal possible number of containers in its TE policies.¶
A small network is divided into areas, depending on the scalability of devices and the routing protocol. Each area is assigned the appropriate number of sets based on its current size. Devices within each area are assigned locators from the sets of the area.¶
Example:¶
Block ID: 5f00:0000::/32¶
Area 1 - 500 devices - 2 sets 0x00 and 0x01¶
First device locator: 5f00:0000:0001::/48¶
Last device Locator: 5f00:0000:01ff::/48¶
A large network is any network with more than approximately 35,000 devices. In this case, a single block is not enough to assign a unique locator to each device. Therefore, the network is divided into regions, with each region able to contain up to about 35,000 devices. The number of regions depends on the total network size.¶
It is important to make each region as large as possible to minimize the number of block boundaries, which benefits efficient TE. Note that a "region" here is a logical grouping and is not limited to a geographical area. Within each region, sets are assigned following the same principles as in small network addressing.¶
For a network with up to approximately 500,000 SRv6 devices, 16 regions are needed. The addressing format is the following:¶
BBBB:BBBR:SSNN::¶
Where:¶
Example:¶
In a network using SRv6 Space 5f00:0000::/28, the locator of a device with Node ID 0x43 in region 0x5, set 0x75, is:¶
5f00:0005:7543::/48¶
For a very large network with more than 500,000 SRv6 devices, more regions are needed. In this case, allocate 8 bits (2 nibbles) for the Region ID, allowing to scale up to approximately 8 million SRv6 devices. The addressing format becomes:¶
BBBB:BBRR:SSNN::¶
Where:¶
Example:¶
In a network using SRv6 Space 5f00:0000::/24, the locator of a device with Node ID 0x43 in region 0x11, set 0x75, is:¶
5f00:0011:7543::/48¶
Using Flexible Algorithms (FAs) in SRv6 networks requires assigning multiple independent locators to single device, one for each algorithm. To ensure efficient summarization schema and operational simplicity, follow these guidelines:¶
Typically, a single nibble (4 bits) is enough for the FA ID.¶
It is very unlikely that a service provider will use more than 16 flexible algorithms.¶
Locator Structure for small networks:¶
BBBB:BBBF:SSNN::¶
Where:¶
Locator Structure for large networks:¶
BBBB:BBFR:SSNN::¶
Where:¶
Locator Structure for very large networks:¶
BBBB:BFRR:SSNN::¶
Where:¶
Example Small Network:¶
In a network using the default algorithm (Algorithm 0) and two flexible algorithms, using SRv6 Space 5f00:0000::/28, the locators of a device with Node ID 0x43 in set 0x75 are:¶
Example Large Network:¶
In a network using the default algorithm and two flexible algorithms, using SRv6 Space 5f00:0000::/24, the locators of a device with Node ID 0x43 in region 0x5, set 0x75 are:¶
The addressing of interfaces and loopbacks is independent of locator addressing. If the already uses legacy IPv6 addressing, there is no need to change it. But some synergic addressing strategies can simplify network operations and increase scalability.¶
Assigning loopback addresses from the node's locator space has several advantages:¶
Although loopback addresses can technically be assigned from any algorithm's locator, it is most logical and common to use the default algorithm's locator.¶
Example:¶
For a device with Node ID 0x43 in set 0x75:¶
Each IPv6 interface has a Link-Local unicast address ([RFC4291]). Link-local addresses are sufficient to establish routing adjacencies and build a fully functional IPv6 and SRv6 network. Therefore, it is possible to use only link-local addresses on infrastructure links between routers ([RFC7404]).¶
This simplifies the overall addressing plan and reduces the IGP database but comes with the drawback that interfaces with only link-local addresses cannot be reached remotely, which may cause operational challenges.¶
An elegant workaround to mitigate this limitation is to assign End.X SIDs to all interfaces. Then the interfaces are remotely pingable since End.X SIDs are remotely reachable over ICMP.¶
This document does not introduce new security issues beyond those existing in SRv6 and IPv6.¶
This document makes no IANA requests.¶
The authors would like to acknowledge the contribution of Francois Clad, Clarence Filsfils and Ketan Talaulikar for their valuable input and review of this document.¶