JN0-664 JNCIP-SP: Advanced BGP, VPNs, and Service Provider Routing

The active professional service-provider routing exam is JN0-664, which leads to the Juniper Networks Certified Professional, Service Provider Routing and Switching credential. Juniper introduced it on December 12, 2022 after retiring JN0-663. As of October 2026, Juniper lists a 90-minute, 65-question Pearson VUE exam based on Junos OS 22.3, with JNCIS-SP as the prerequisite certification for the professional credential.

The active blueprint covers OSPF, IS-IS, BGP, class of service, IP multicast, Layer 3 VPNs, and multiple Layer 2 VPN technologies including BGP Layer 2 VPNs, LDP Layer 2 circuits, VPLS, and EVPN. Candidates preparing now can use JN0-364 as the current specialist context and JN0-663 only as the immediate historical predecessor.

The challenge of JN0-664 is not any single protocol. It is the number of dependencies that must be kept straight at once. Provider underlay reachability supports BGP and label transport; BGP distributes both ordinary and service-specific routes; route targets control VPN membership; labels map customer traffic through the core; multicast and CoS add their own state. Preparation should mirror that layered architecture.

Start every service-provider problem with transport reachability

Before a VPN or advanced service can work, provider infrastructure addresses must be reachable. OSPF or IS-IS commonly provides that base. A professional candidate should know how link-state information floods, how metrics influence path selection, how areas or levels affect propagation, and how routing policy can summarize, leak, or restrict information. The goal is a stable underlay that higher-level services can trust.

In the lab, verify loopback reachability before adding BGP or MPLS-dependent services. Then change one metric or introduce one adjacency fault and record what breaks upstream. This teaches the dependency chain. If a remote PE loopback is unreachable, a VPN failure is not primarily a VPN problem. Knowing which layer must be healthy first is one of the most important troubleshooting habits in service-provider operations.

OSPF and IS-IS demand protocol-specific evidence

OSPF and IS-IS are both link-state protocols, but their operational vocabulary and failure modes are not identical. JN0-664 expects candidates to understand OSPF area types, LSA behavior, DR and BDR operation, metrics, summarization, virtual links, and OSPFv2 versus OSPFv3. It also covers IS-IS areas and levels, LSP flooding, DIS behavior, wide metrics, summarization, and route leaking.

Study each protocol with its native database and adjacency tools. When a route is missing, ask whether the relevant topology information exists, whether it propagated to the right scope, and whether policy or preference prevented installation. This is faster than treating a missing route as a generic problem. Professional skill means knowing which protocol artifact should prove each part of the control-plane story.

BGP scaling is as important as BGP route selection

Service providers use BGP at a scale where a full mesh of peers is often impractical. Route reflectors, multipath, communities, advanced policy, and multiple address families become normal design tools. JN0-664 also includes BGP FlowSpec and route damping concepts, reinforcing that BGP can distribute more than ordinary reachability and can influence operational responses to network conditions.

Create a route-reflector lab with at least two client groups. Observe what routes are reflected, what next-hop values remain, and how policy affects propagation. Add communities and use them to control export. Then enable a second address family and verify that the session carries separate kinds of information. This provides the mental flexibility needed for VPN and EVPN topics, where BGP transports service state rather than only Internet prefixes.

Class of service turns capacity into differentiated service

Provider networks carry customers and applications with different expectations. Junos CoS uses classifiers, forwarding classes, loss priorities, policers, schedulers, drop profiles, and rewrite rules to decide how packets are treated when resources are constrained. Candidates need to understand the processing sequence and how behavior changes from ingress classification through egress scheduling.

Build a congestion test deliberately. Mark two or three traffic classes, send enough traffic to exceed an interface’s capacity, and inspect queue statistics. Determine which class is delayed or dropped first and why. Then change one scheduler or policer and repeat. This turns CoS from a list of configuration knobs into an observable service policy, which is how it must be understood in production.

Multicast requires proving the tree, not just the route

IP multicast adds receiver state and tree construction to ordinary routing. JN0-664 covers ASM, SSM, RPF, IGMP, PIM dense and sparse modes, rendezvous points, Anycast RP, and routing policy. A receiver can have correct unicast reachability and still receive no multicast if membership, PIM state, RPF, or source discovery is wrong.

Trace the multicast service from the receiver backward. Confirm IGMP membership, the outgoing-interface list, upstream PIM state, RPF neighbor, and source or rendezvous-point information. For SSM, account for the source-specific join behavior. By validating the tree stage by stage, candidates learn to diagnose multicast as a control-plane construction problem rather than a mysterious form of packet loss.

Layer 3 VPNs combine BGP, VRFs, labels, and policy

An MPLS Layer 3 VPN works because multiple mechanisms align. Customer routes enter a VRF, route distinguishers make overlapping prefixes unique, route targets decide which remote VRFs import them, MP-BGP distributes the VPN routes, and labels carry traffic across the provider core. Site of origin, sham links, table labels, inter-provider models, and Internet-access options add further complexity.

Choose one customer prefix and document its lifecycle. Where is it learned? Which RD is attached? Which RTs control membership? What label is advertised? Which remote PE installs it, and into which VRF? What transport label reaches that PE? If the service fails, this sequence gives a structured troubleshooting path. It also makes clear why a route can exist in BGP yet still fail to reach the intended customer table.

Layer 2 VPN technologies solve similar customer needs with different control planes

JN0-664 includes BGP Layer 2 VPNs, LDP Layer 2 circuits, VPLS, and EVPN. All can provide Ethernet-oriented services, but they differ in discovery, signaling, learning, scaling, and multihoming behavior. Candidates should be able to identify which control plane is responsible for establishing a service and which tables or labels prove that it is operational.

Build a comparison matrix from the customer’s perspective and the provider’s perspective. The customer may simply expect two sites to share Layer 2 connectivity. The provider must know whether LDP creates a pseudowire, BGP carries Layer 2 NLRI, VPLS learns MAC addresses across a multipoint service, or EVPN distributes MAC information in BGP. Understanding those distinctions is far more useful than memorizing a list of protocol names.

EVPN shows how service-provider control planes are converging

EVPN is important because it combines BGP signaling with Ethernet services and robust multihoming. In the JN0-664 context, candidates should understand traffic flow, MAC learning and distribution, label signaling, and multihoming behavior. The same conceptual family now appears in data center fabrics, which makes EVPN one of the most transferable technologies in advanced Juniper certification.

Study one multihomed site and one remote site. Determine how endpoint reachability is advertised, how redundancy information is represented, and what happens when one attachment fails. Then trace a frame through the provider. This bridges the gap between abstract BGP route information and the actual Ethernet service a customer consumes.

Preparation should move from a clean underlay toward complete services

The most productive JN0-664 lab sequence is cumulative. Build OSPF or IS-IS and verify infrastructure reachability. Add BGP and route policy. Introduce the transport and then one Layer 3 or Layer 2 VPN. Add multicast or CoS only after the base service works. Finally, break one dependency and troubleshoot from the lowest failed layer upward. This mirrors how provider networks are actually assembled and operated.

Candidates who need a fundamentals refresh can use JN0-106 and IPv4 subnetting before spending time on advanced VPN behavior. The wider Juniper certifications portfolio also shows how service-provider skills overlap with enterprise and data center routing. JN0-664 rewards depth, but that depth is built on reliable fundamentals and a disciplined method for proving control-plane and forwarding state.

Time spent on route tables should be paired with time spent on forwarding tables and label state. The control plane may select the expected route while the data plane is missing the corresponding label or next-hop information. Conversely, stale forwarding state can create symptoms after the routing protocol has already converged. Professional troubleshooting means checking the layer that actually carries the packet, not assuming the routing table alone proves service health.

VPN labs are stronger when they use overlapping customer prefixes. Overlap forces the candidate to understand why VRFs, route distinguishers, and route targets exist instead of treating them as ceremonial configuration. Add two customers that both use the same private prefix, then verify that their routes remain distinct and that only intended sites exchange information. The exercise makes isolation visible and memorable.

Service-provider engineers also need disciplined change control. Before altering policy, labels, multicast, or CoS, identify the affected customers and the rollback condition. After the change, validate both the target service and neighboring services that share infrastructure. This operational habit is not a separate exam topic, but it is the context in which every JN0-664 technology is used.

JN0-664 is a professional routing exam in the fullest sense: it asks candidates to reason across underlay protocols, BGP policy and scaling, multicast, traffic treatment, and customer VPN services. The individual technologies matter, but the real difficulty is understanding how they depend on one another.

Prepare by building complete but manageable service topologies and validating each layer before moving upward. When something breaks, identify the lowest failed dependency and prove it with protocol-specific evidence. That habit makes the current blueprint far more coherent and develops exactly the kind of troubleshooting discipline expected from a JNCIP-SP engineer.