Segment Routing (SRv6)

SRv6 encodes the entire forwarding path as an ordered list of segments (SIDs) within an IPv6 Segment Routing Header (SRH). The ingress node defines the path, eliminating per-flow state in the core. Each SID is a 128-bit IPv6 address representing a topological or service instruction.

• Topological SIDs: Direct traffic through specific nodes or links
• Service SIDs: Invoke functions like deep packet inspection or caching proxies
• End.B6.Encaps: A behavior that steers packets into a specific SRv6 policy

SRv6 enables hard isolation between network slices by binding each slice to a unique locator prefix within the SID space. A single physical infrastructure supports multiple logical networks with distinct QoS guarantees, routing policies, and cache node assignments.

• Slice-Aware Caching: Steer content requests through slice-specific cache hierarchies
• Dedicated SID Spaces: Prevent cross-slice interference at the forwarding plane
• Flex-Algo Integration: Combine SRv6 with IGP Flexible Algorithms for per-slice topology computation

SRv6 treats network functions as segments, enabling dynamic service function chaining without protocol overhead. A packet can traverse a sequence of cache nodes, firewalls, and load balancers defined entirely by its SID list.

• SR-aware Cache Nodes: Advertise themselves as service SIDs for content retrieval
• Dynamic Re-chaining: Modify the SID list mid-path based on cache hit/miss events
• Stateless Processing: Each node executes its function independently using the SRH

SRv6 policies enable explicit traffic engineering to direct content requests toward optimal cache nodes based on latency, load, and content popularity predictions. The controller computes paths that minimize backhaul utilization.

• Low-Latency Paths: Route delay-sensitive content through edge caches with sub-millisecond RTT
• Load-Aware Steering: Distribute requests across cache clusters using weighted SID lists
• TI-LFA Protection: Sub-50ms failover to backup cache nodes using Topology-Independent Loop-Free Alternates

The SRv6 Network Programming framework defines behaviors bound to SIDs, creating a programmable data plane. End.AN (Auto-Next) and End.DX6 (Decapsulation and IPv6 Cross-Connect) behaviors enable precise cache node selection and traffic termination.

• End.DT6: Decapsulate and forward to a specific IPv6 table for cache isolation
• End.B6.Insert.Red: Insert a reduced SRH for bandwidth-constrained links
• uSID Compression: Compress SID lists to reduce header overhead in cache-bound traffic

SRv6 enables federated caching across administrative domains by using globally routable SIDs. Different operators or enterprise tenants can share cache infrastructure while maintaining strict traffic isolation through slice-specific SID namespaces.

• Inter-Domain SIDs: Extend cache steering across AS boundaries without tunneling overlays
• Tenant-Specific Cache Policies: Assign dedicated cache node SIDs per enterprise slice
• BGP-LS Advertisement: Distribute cache node SIDs and capabilities via link-state routing protocols

A virtualization architecture that partitions a single physical network into multiple isolated end-to-end logical networks. Each slice can have its own topology, security, and Quality of Service (QoS) rules.

• Hard Slicing: Dedicated physical resources with strict isolation
• Soft Slicing: Shared resources with logical separation using VPNs
• SRv6 uses the SID (Segment Identifier) to encode slice membership directly in the packet header, binding a traffic flow to a specific cache node chain.

A routing paradigm where the ingress node specifies the complete path a packet will take through the network. Intermediate nodes simply execute the instructions encoded in the packet header.

• Eliminates the need for per-flow state on transit routers
• SRv6 uses an IPv6 Segment Routing Header (SRH) to carry an ordered list of segments
• Enables deterministic traffic steering through specific MEC Caching nodes based on slice requirements

A caching strategy where the decision of which cache node serves a request is determined by the network slice the user belongs to. Different slices can be steered to different cache tiers.

• A URLLC slice may be routed to a local base station cache for sub-millisecond latency
• An eMBB slice may be directed to a regional MEC cache with higher capacity
• SRv6 encodes this policy as a Service Function Chain (SFC) in the packet's segment list

A 128-bit Segment Identifier that combines location, function, and arguments into a single IPv6 address. The SID tells a node where to forward and what to do.

• LOC:FUNCT:ARGS format: Locator identifies the node, Function specifies the action (e.g., End.Cache), and Arguments carry metadata
• The End.Cache function instructs a node to serve content from its local Content Store
• Enables Programmable Data Planes where caching behavior is defined by the SID itself

An SR Policy defines a specific forwarding path and its associated Service Level Agreement (SLA) requirements. It binds a traffic flow to a segment list.

• Policies can be explicit (statically defined) or dynamic (computed by a controller)
• A policy can specify low-latency paths through cache nodes for video streaming slices
• Uses headend steering where the ingress PE router applies the policy based on packet classification

A network architecture where content is addressed by name rather than host location. SRv6 can bridge ICN principles with IP routing.

• Named Data Networking (NDN) uses Interest and Data packets with a Content Store at each node
• SRv6 can encode an ICN name as a SID, steering an Interest packet to the nearest cache holding the named content
• Enables native in-network caching without overlay protocols, reducing backhaul load