Revisiting the Control Plane and Data Plane
Every networking device covered throughout this series conceptually performs two distinct functions. The Control Plane makes decisions -- running routing protocols like OSPF and BGP, discussed earlier in this series, to determine the best path to every destination. The Data Plane (also called the forwarding plane) actually moves traffic based on those decisions -- the CEF-based forwarding, discussed earlier in this series, that switches and routers perform on every packet.
Traditional (distributed) architecture:
Every single router and switch runs its own
control plane AND its own data plane locally --
each device independently calculates its own
routing table and independently forwards traffic
based on its own local decisionsThe SDN Shift: Centralizing the Control Plane
Software-Defined Networking (SDN) separates these two functions architecturally: a centralized SDN Controller handles the control plane decision-making for an entire fleet of devices, while the individual devices retain only the data plane function of actually forwarding traffic based on instructions received from the controller.
SDN (centralized) architecture:
SDN Controller: makes all forwarding decisions
for the entire network, maintaining a
network-wide view rather than each device
independently calculating its own limited view
Network Devices: retain only data plane
functionality, forwarding traffic exactly
as instructed by the controller, without
independently running routing protocol
calculations themselvesThis is conceptually the same architectural pattern already seen twice earlier in this series -- the centralized WLC managing lightweight access points for wireless networks, and controller-based networking managing traditional wired infrastructure -- now formalized as SDN's core architectural principle and applied across an entire network fabric.
Why Centralization Provides Real Benefits
Benefits of centralized control:
Global network visibility: the controller sees
the entire network's state at once, rather
than each device only knowing about its
immediate neighbors, as with OSPF or EIGRP,
discussed earlier in this series
Simplified device configuration: individual
devices become comparatively simple forwarding
engines, with complex policy logic centralized
at the controller rather than distributed
and duplicated across every device
Faster, more consistent policy changes: a policy
update made once at the controller propagates
consistently to every affected device, rather
than requiring the same change to be manually
applied device by device, as with the
traditional CLI configuration covered
throughout most of this seriesSD-WAN: Applying SDN Principles to Wide Area Networks
SD-WAN (Software-Defined Wide Area Network) applies the same control plane centralization to WAN connectivity, discussed earlier in this series, specifically addressing the limitations of traditional WAN designs that relied entirely on expensive, dedicated circuits like MPLS.
Traditional WAN design:
Branch offices connect to a central location
using dedicated, expensive circuits (MPLS),
with limited ability to dynamically choose
between different paths based on real-time
conditions
SD-WAN design:
Branch offices connect using ANY available
transport -- MPLS, broadband internet, LTE/5G --
with a centralized SD-WAN controller
dynamically selecting the best path for each
type of traffic based on real-time link
quality measurementsKey SD-WAN Capabilities
Application-aware routing: different applications
can be routed over different underlying transport
links based on their specific requirements --
voice traffic, requiring the low latency
discussed earlier in this series regarding QoS,
might use a dedicated MPLS link, while bulk
data transfer uses cheaper broadband internet
Dynamic path selection: the SD-WAN controller
continuously monitors each available path's
actual performance (latency, jitter, packet
loss) and can shift traffic away from a
degrading path in near real-time, without
waiting for a link to completely fail the
way traditional routing protocol convergence,
discussed earlier in this series, requires
Zero-touch provisioning: a new branch device
can be shipped to a remote site, plugged in,
and automatically receive its full configuration
from the central controller, without requiring
a network engineer to manually configure it
via CLI on-site or remotelyWhy SD-WAN Matters for Modern Enterprise Connectivity
The shift toward cloud-hosted applications has fundamentally changed WAN traffic patterns: traffic increasingly flows directly from a branch office to a cloud provider rather than being routed back through a central data center first, a pattern traditional hub-and-spoke MPLS WAN designs handle poorly, since it forces cloud-bound traffic through an unnecessary detour. SD-WAN's ability to intelligently route traffic directly toward its actual destination, using whichever available transport link is currently best suited for that specific traffic type, directly addresses this shift.
The Trade-Offs of Centralized Control
Considerations when adopting SDN/SD-WAN:
Controller as a critical dependency: since the
controller makes the actual forwarding
decisions, its availability and redundancy
become critical -- most production SDN and
SD-WAN deployments require carefully designed
controller redundancy, unlike a traditional
distributed control plane where no single
device's failure affects the entire network's
ability to make routing decisions
Requires new operational skills: administrators
need to learn the specific controller platform's
interface and abstractions, which may differ
significantly from traditional CLI-based
device-by-device configuration covered
throughout most of this seriesWhy This Represents a Genuine Architectural Shift
SDN and SD-WAN do not replace the underlying networking concepts covered throughout this entire series -- routing, switching, and QoS still fundamentally determine how traffic actually moves through a network. What changes is where the intelligence controlling those decisions lives: rather than being distributed and independently calculated by every device using protocols like OSPF and EIGRP, discussed earlier in this series, that intelligence becomes centralized, giving network operators a single point of visibility and control over decisions that were previously scattered across hundreds or thousands of independently operating devices.