Internet do Futuro - Seminário sobre SDN e NFV

Summary

The presentation explains how Software-Defined Networking (SDN) and Network Functions Virtualization (NFV) can make networks more programmable and adaptable, while also introducing new security and reliability risks.

Traditional networks and their limitations

In a traditional network, each device combines the control plane—the logic that decides how traffic should be handled—with the data plane, which forwards traffic. As networks grow, this arrangement can make them harder to manage because:

• Configuration often depends on each vendor’s equipment and features.
• Manual configuration takes time and increases the chance of mistakes.
• The same policy may be applied inconsistently across devices.
• Responding to failures or sudden traffic increases may require manual intervention.

How SDN changes network control

SDN separates the control plane from the data plane:

• Switches forward traffic according to rules.
• A controller holds the network’s control logic and has a broader view of the network.
• Applications request or define network behavior.
• Applications communicate with the controller through a northbound API, while the controller communicates with switches through a southbound API.

This structure can support greater automation, a more consistent network-wide view, and faster responses. However, centralizing control also concentrates risk: a controller failure can disrupt the handling of new traffic, a compromise can affect a large part of the network, and a faulty policy may be rapidly distributed to many devices.

Technologies and NFV

The presentation describes several technologies and concepts:

• OpenFlow can install or modify forwarding rules on switches.
• P4 allows more control over how switches process packets.
• ONOS and OpenDaylight are examples of SDN controllers.
• NFV implements network functions—such as firewalls, intrusion-detection systems, and load balancers—as software-based VNFs.
• VNFs run on NFV Infrastructure (NFVI), using virtualization or container technologies.
• Service chaining directs traffic through network functions in a specified order—for example, firewall, intrusion detection, then load balancing.

In short, SDN determines where traffic goes, while NFV provides software functions that process it. NFV management and orchestration components coordinate the functions, their lifecycles, and the infrastructure they use. The transcript refers to these roles as NFVO, VNFM, and VIM; one term is likely mistranscribed in the subtitles.

Example: virtualized 5G cloud network

A cloud provider could use SDN and NFV to:

• Create network slices on demand.
• Automate service chains.
• Respond quickly to outages or traffic spikes.
• Use programmable switches and server platforms such as KVM or Kubernetes.
• Implement 5G core and other network functions in software.

In this setup, applications make decisions, the SDN controller programs traffic-flow rules, the NFV orchestrator manages virtual network functions and resources, and the data plane forwards traffic.

Limitations and risks

The presentation highlights four challenges:

1. Scalability: A controller can become a bottleneck. Clustering can help but adds complexity.
2. Security: APIs and controllers can be vulnerable and need strong protection.
3. Resilience: Multiple controllers do not eliminate communication failures or network partitions; switches should keep operating with existing rules where possible.
4. Integration: Real networks combine legacy and newer equipment from different vendors. Teams also need skills in programming, automation, and security.

Measures to reduce risk

The presenter recommends:

• Using a cluster of controllers rather than relying on one node.
• Protecting communication channels with measures such as TLS and authentication, and controlling access to the relevant interfaces.
• Ensuring switches can continue using existing rules during temporary controller outages.
• Using multi-domain architectures to limit how far a failure can spread.
• Testing policies in an emulation environment before deploying them on a production network.
• Maintaining recovery mechanisms.

Main conclusion

Programmable networks can be both more capable and more vulnerable. SDN and NFV enable automation, network partitioning, service chaining, faster reactions, and a broader view of network behavior. But poor design can turn controllers, APIs, software bugs, or incorrect policies into critical risks.

The key issue is not centralization by itself, but how it is designed—with redundancy, security, and recovery in mind. SDN’s centralization is primarily logical and does not necessarily require dependence on a single physical server.

Speakers and sources featured

• Speaker: Zac, the presenter (as identified in the subtitles).
• Other speakers: None identified.
• Sources: The presenter refers to a final slide of references, but the subtitles do not provide the reference titles or authors. Technologies and systems mentioned include OpenFlow, P4, ONOS, OpenDaylight, NFV, KVM, and Kubernetes.