Any IT professional looking at software-defined networking (SDN) implementation will need to have a long and in-depth discussion about network virtualization (NV) with management tool vendors. Enterprise Management Associates (EMA) research has found that the majority of early adopters have management challenges with NV and SDN, as their standing network management tools do not fully support a virtualized environment.
Many of the IT professionals surveyed are modifying the tools they have to work in a virtualized environment, while many others are acquiring new tools. Based on a survey of early SDN adopters (150 enterprise IT professionals and 76 communication service provider infrastructure professionals), EMA’s report “Managing Tomorrow’s Networks: The Impacts of SDN and Network Virtualization on Network Management” reveals the challenges and benefits of SDN adoption.
Challenges With Both Overlays and Underlays
In the data center, users see a more significant challenge for NV overlays (VMware NSX, Nuage Networks, etc.) than for NV underlays like Cisco ACI or OpenFlow-based solutions from the likes of Big Switch Networks or NEC. For instance, 37 percent of early enterprise adopters said their network planning and engineering tools fully support underlays. While that percentage is by no means encouraging, it is higher than the 30 percent who said their planning and engineering tools fully support overlays. Furthermore, 40 percent of research participants said their performance monitoring tools fully support underlays, but just 33 percent said these tools are ready for overlays.
While both flavors of data center SDN and NV pose challenges for network management systems, overlays pose the bigger challenge. We asked early adopters to name their top overlay management challenges with SDN. The top two challenges were “data and control plane visibility” (37%) and “end-to-end monitoring across physical and virtual network elements” (35%). The third biggest overlay challenge is “end-to-end troubleshooting across physical and virtual network elements” (32%). The first challenge is relevant to any kind of SDN implementation, but naysayers identified end-to-end visibility and troubleshooting as a potential problem before people started adopting overlays.
Clearly, early adopters have encountered the management challenges with NV and SDN that detractors predicted. To rectify the problem, they need their management tool vendors to evolve. Network management tools are rooted in a hardware-centric paradigm. They are designed to support boxes, not software elements. But overlays and underlays both change how networks are built and operated. In my conversations with network management vendors, I have found a mix of positions on this issue. Some management tool vendors are updating their products today. Others have SDN roadmaps. But some are still taking a wait-and-see approach. If they have plans for SDN support, they haven’t shared them with me yet.
Planning and Engineering Top List
How should network management tools evolve? EMA posed this question to early enterprise adopters across four tool categories: planning and engineering, availability monitoring, performance monitoring, and troubleshooting. There is too much data to cover in this one post, but I’ll highlight a couple leading SDN requirements in planning/engineering and performance monitoring.
Early enterprise adopters identified two top SDN requirements of planning and engineering tools: “Network state analytics and simulation” (48%) and “capacity planning based on SDN flows.” The software-driven nature of SDN makes networks extremely dynamic, which in turn makes it far more challenging for network architects and engineers to model network state. In planning and engineering tasks, engineers need visibility into network state in order to right-size capacity for applications and services. They likewise need tools that allow them to plan capacity based on SDN flows.
Speaking of the dynamic nature of SDN, the top SDN requirement for performance monitoring tools is the ability to “understand and adjust monitoring of on-demand capacity changes” (47%). Many performance monitoring solutions have static probes where sensors are deployed. They collect packets or network flows from strategic points on the network. But when SDN implements capacity changes programmatically, the critical points for network data collection can shift without warning. A monitoring tool needs to react to these changes.
As I mentioned above, EMA’s full report on the subject of SDN management contains an immense amount of data on the readiness of network management tools for SDN. We examined, not only data center SDN technologies, but SD-WAN, campus SDN, and enterprise NFV as well. We also polled service provider SDN adopters on their management challenges with NV. The research also explores SDN use cases, implementation challenges, and business benefits. If you’d like to learn more, download our full report.