When a new application or service is installed in a traditional network, typically a planning group gets involved to assess whether or not the network is ready for it. With SDN, engineers no longer get weeks to prepare for the rollout of a new application. In fact, programmability takes the operator out of the equation.
In addition, traditional manual and device-centric management methods cannot provide the visibility needed to run a programmable network that automatically adapts to application demands. This is especially true across the dynamic, resource-constrained WAN where management is more complex than in the datacenter.
If applications and services are being rolled out without operator intervention and adequate visibility, how do you plan for them? Who or what governs whether or not these programmatic changes should be made? How do you know if the network can support a new request without negatively impacting existing applications?
Many think that if you remove error-prone humans from network planning and operations, problems won’t happen. But people do not cause many fault and performance issues. SDNs are still prone to link or node failure, software bugs, unforeseen protocol interactions, and other issues.
In addition, SDN presents brand-new challenges. For example, in a large WAN, how many controllers should you have? Where should you place them? What if two controllers talking to the same switch give conflicting instructions?
Arbitrating competing network resource requests will be critical as will setting up new paths in the first place. You will need to verify that paths are set up properly and conduct ongoing monitoring. Things like visualization of the current SDN topology will be critical, so you can understand the impact of programmatic changes.
Automating Network Management
The main challenge is being able to move today’s management practices into the automation realm. We must be able to understand what impact an application requesting resources from an SDN will have on performance. SDN creates the need to replicate traditional functions of capacity planning, monitoring, troubleshooting, security, and other critical management capabilities.
It’s less a question of needing to develop brand new ways to manage networks, but more about adapting what we know are good, robust network management techniques.
For example, route analytics is proven to improve availability and performance in the traditional network with real-time, network-wide visibility into topology details, routing events and traffic patterns. This is directly applicable to SDN, because successfully monitoring and managing SDN applications requires always-current network models and traffic load profiles as well as the ability to predict the impact of change to network routing topologies and traffic flows.
[Read why Tom Hollingsworth thinks the debates over OpenFlow, NSX and ACI miss the point in "Focus On SDN Tools Obscures SDN Benefits."]
In addition, route analytics can take advantage of the unique application insight found in SDNs, with network diagnosis, analysis and reporting that is application-aware. For instance, it can report whether the application's servers are best positioned to serve that application's user base.
All this requires more detailed and sophisticated analysis of what’s happening in the network. Vendors and the SDN community at large are mostly ignoring SDN analytics, which are crucial for:
• Troubleshooting and visualization.
• Determining the network state at any given time.
• Inspecting and replaying events learned both from the controller and the network devices.
• Comparing routing state and paths when a service/application is performing well and when it is not.
• Monitoring paths for changes in hops, metric, delay, and bandwidth.
An awareness of the management challenges of SDN is crucial, but then the industry needs the right tools to move management into the automation realm. Network professionals need the same level of visibility and control into SDNs as they do in the traditional network. This requires some new technologies as well as adaptions of existing technologies and best practices that provide the intelligence and analytics needed to address the unique challenges of software defined networking.
Only when organizations have adequate management will the promise of SDN to be transformational and disruptive be realized.
Cengiz Alaettinoglu is CTO of Packet Design.