
Non-Stop Forwarding (NSF): How It Keeps Networks Running
Networks are expected to run without interruption, even when hardware fails, software restarts or a control component is replaced. Non-stop forwarding, or NSF, is a high-availability technology that helps routers keep forwarding packets during a control-plane failure, so users experience little or no disruption. It is one of the quiet mechanisms behind the reliability of enterprise, service-provider and data-center networks.
This guide explains what non-stop forwarding is, how it works with related technologies, where it is used, and what network teams need to consider when deploying it.
What Is Non-Stop Forwarding?
In a router, two planes do different jobs. The control plane runs routing protocols, builds the routing table and decides where packets should go. The forwarding plane, or data plane, actually moves packets according to that table. Normally, if the control plane fails, for example when a routing engine restarts or fails over to a backup, the forwarding table could be lost and traffic would stop until routing reconverges.
Non-stop forwarding prevents that interruption. With NSF, the forwarding plane continues to forward packets using the last known forwarding information while the control plane recovers. Once the control plane is back, it relearns routing information from neighbors without dropping the traffic that kept flowing. The result is that a control-plane event, which might otherwise cause an outage of seconds or minutes, becomes invisible or barely noticeable to users.
How NSF Works
NSF relies on separation between control and forwarding, and on cooperation from neighboring routers. When a router with NSF restarts its control plane, it signals its neighbors that it is restarting gracefully, using extensions to routing protocols such as OSPF, IS-IS and BGP, known as graceful restart. Neighbors keep their routes to the restarting router and continue sending it traffic, rather than immediately declaring it down and rerouting everything.
Meanwhile, the forwarding plane keeps using its existing forwarding table. When the control plane comes back, it rebuilds its routing table from neighbor information and synchronizes it with the forwarding plane. If the rebuild succeeds within the graceful-restart window, traffic continues without a break. If it takes too long, neighbors eventually treat the router as failed and reroute, so NSF improves resilience but does not make a router indestructible.
NSF is often paired with stateful switchover (SSO), which is the ability of a device with redundant control processors to switch from an active to a standby processor while preserving state. SSO provides the redundant control hardware, while NSF provides the forwarding continuity during the switchover. Together, they are a standard high-availability design for core and distribution routers.
NSF vs. Related High-Availability Technologies
| Technology | What It Does | Main Limitation |
|---|---|---|
| Non-stop forwarding (NSF) | Keeps forwarding during control-plane restart | Needs neighbor cooperation and timely recovery |
| Stateful switchover (SSO) | Switches control to a standby processor with state intact | Requires redundant hardware in the device |
| Graceful restart | Protocol extensions that let neighbors help a restarting router | Must be supported by neighbors |
| Redundant links / ECMP | Provides alternate paths if a link fails | Does not by itself protect the control plane |
| Virtual router redundancy (VRRP/HSRP) | Provides a backup gateway if one router fails | Gateway failover can still cause brief disruption |
These technologies are complementary, not alternatives. A resilient network typically combines NSF and SSO inside key routers, redundant links between them, and gateway redundancy at the edge, so that no single failure, whether in hardware, software or a link, stops traffic for long.
Where NSF Is Used
NSF matters most where downtime is costly or dangerous: service-provider cores, enterprise data-center interconnects, financial trading networks, healthcare systems and government networks. In these environments, even a brief routing outage can disrupt transactions, calls, video or critical applications. NSF is also important in networks that use frequent software upgrades, because it allows a control-plane restart for an upgrade to occur without dropping user traffic, when combined with in-service software upgrade techniques.
It is less critical for small, simple networks where a brief outage is acceptable and where routers may not support NSF or have redundant control planes. In those cases, simpler redundancy, such as dual internet links or a backup router, may be sufficient and more cost-effective.
Security and Operational Considerations
High availability and security must be designed together. A router that keeps forwarding during a restart must still enforce access controls, firewall policies and routing security. Teams should verify that forwarding during NSF does not bypass security policies, that routing protocol authentication is enabled so that graceful-restart signals cannot be spoofed by an attacker, and that logs capture control-plane events for investigation.
Operationally, NSF requires testing. A feature that is configured but never tested may fail when it is needed most. Test control-plane failover in a maintenance window, measure how long recovery takes, check that neighbors support graceful restart, and document the behavior for each router model and software version in use. Keep software versions consistent where possible, because NSF behavior can vary between releases.
Testing and Monitoring NSF
A reliable NSF deployment is proven, not assumed. Start in a lab or a non-production segment that mirrors your routing design, then force a control-plane switchover and measure packet loss, convergence time and application impact. Record whether neighbors held their routes, whether the forwarding table remained stable, and how long the control plane took to resynchronize. Repeat the test after every major software upgrade, because protocol behavior and timer defaults can change between releases and silently weaken a previously working design.
In production, monitor control-plane restarts, graceful-restart events and forwarding-plane errors as closely as you monitor link status. Alert on unexpected restarts, on neighbors that do not support graceful restart, and on any mismatch between the routing table and the forwarding table after recovery. These signals often reveal a failing supervisor, a memory problem or a configuration drift long before users notice an outage, giving engineers time to fix the root cause during a planned window.
Deploying NSF: Practical Steps
First, inventory your routers and confirm which models support NSF, SSO and graceful restart for the protocols you run. Second, enable the features consistently on paired routers and their neighbors, because a one-sided configuration will not work as intended. Third, set graceful-restart timers appropriately for your network size and convergence time. Fourth, monitor control-plane events and forwarding continuity, and alert on unexpected restarts. Finally, include NSF failover in your disaster-recovery and change-management tests, so that upgrades and failovers are routine rather than risky.
Network continuity also depends on the wider infrastructure. For the facilities that house core routers, see this overview of what data centers are and how they work. Perimeter controls that protect those networks are explained in this guide to what a network firewall does, and for performance under load, read about the application acceleration manager and how it speeds up apps.
Frequently Asked Questions
What is non-stop forwarding?
Non-stop forwarding is a router technology that keeps forwarding packets using existing forwarding information while the control plane restarts or fails over, avoiding a traffic outage.
How is NSF different from SSO?
SSO switches control to a standby processor while preserving state, while NSF keeps the forwarding plane running during that switch or a control-plane restart; they are usually used together.
Does NSF require support from neighboring routers?
Yes. Neighbors must support graceful restart and keep their routes during the restart, otherwise they will reroute traffic and the benefit of NSF is lost.
Does NSF protect against link failures?
Not directly. NSF protects against control-plane failures; link failures are handled by redundant links, fast reroute and routing reconvergence.
Do small networks need NSF?
Usually not. NSF is most valuable in core, provider and data-center networks where brief outages are costly; smaller networks can often rely on simpler redundancy.



[…] forwarding should not collapse while BGP reconverges. Related resilience techniques help, such as non-stop forwarding, which keeps packets flowing during a control-plane restart. When traffic shifts unexpectedly, […]