Executive Summary: Software-Defined Networking in the Data Center
All IT organizations deal with complexity. Having complexity itself is not the result of mediocracy . In fact, it’s the opposite. Complexity is a byproduct of success . Complexity can be an artifact of rapid growth or the result of mergers and acquisitions. Yet regardless of its origins, complexity is a problem. It’s the enemy of efficiency, security, and innovation.
The network is next. Not because physical networking is going away . There will always be physical networking infrastructure to move bits and bytes from A to B, deal with latency, manage flows, and more. Physical network infrastructure continues to evolve and become rock solid based on the specific needs of data center fabrics, campus networks, and routed WANs. Hardware forms a web of highly reliable connections, which by itself is a complex task to refine to a highly efficient operational model.
But history has taught us that when we rely on hardware to do more complicated things, like defining policies that automatically follow dynamic ephemeral workloads, hardware starts to buckle under the pressure . Further, the cloud has no hardware boundaries and extends transparently through private data centers and on to users.
When you virtualize the network and move to a software-first strategy, two things happen:
- You simplify and make your infrastructure fundamentally more efficient.
- You make your infrastructure more secure, because there are fewer “seams” to exploit.
The Changing State of the Network and Software-Defined Networking
The age-old mantra of “You can’t know where you’re going until you know where you’ve been” holds true for today’s modern data center networking. As we delve into why the world is changing, and how leading with software is becoming more and more critical, it is important to know the recent history of networking to understand how we got to the current state we are in.
The Evolution of Software-Defined Networking
The Open Networking Foundation (ONF) defines software-defined networking (SDN) as “the physical separation of the network control plane from the forwarding plane, and where a control plane controls several devices.” This is a design principle being widely used now but is not an actual solution to specific use cases. While different use cases still use this basic design principle, the problems they address are very different. The following diagram shows the high-level use cases and the market adoption that drove them.
Multi-Cloud Networking: Virtualization paved the way to the private cloud, which demanded the same agility from the network as from the rest of the data center, leading to the first use case of network virtualization to complete the automation needed for the software-defined data center (SDDC). Network virtualization abstracted the networking done for virtual machines (VMs) from that of the physical network. The new network edge had moved into the virtual with an estimated 80M virtual ports and counting. At the same time, public clouds started to become a reality, making it even more critical for a software abstracted network enabling the same operational practices across disparate environments . Further, applications began moving to new micro-service architectures based on containers, leading the movement to cloud-native applications. This shift proved that virtual networking could extend beyond the virtual machine to new future workloads.
Software-First Requirements for Network Upgrades
As your organization approaches network upgrades and the shift to software-defined solutions, it’s important to remember that business drivers should guide infrastructure decisions. The drivers can take the form of higher-level goals to modernize the data center, cloud-first strategies, or something more tactical, such as an in-progress data center consolidation. Either way, the conversation should begin at a strategic level that considers the business and its priorities.
After that, IT practitioners build an environment that supports the new initiative using a software-defined networking strategy aligned with business priorities. For the infrastructure, the high-level requirement of multi-cloud networking takes the form of network virtualization, automation, multi-cloud extensions, and cloud-native approaches. Network security becomes intrinsic to the workload and applications. Fabric management moves to operational efficiency initiatives.
Software-First Design Guidelines
To this point, we have reviewed the requirements and seen the different approaches to SDN. As outlined in the “Software-First Requirements for Network Upgrades” section of this document, not all solutions that follow the principles of SDN are accomplishing the same task. Some solutions may even be provided by different vendors. This section includes a set of design guidelines for upgrading from a traditional networking architecture to an architecture that is automated, works with developers, and contains intrinsic security and cloud. These are all among the benefits of VMware NSX® solutions, which provide a network virtualization platform for the software-defined data center (SDDC).
VMware NSX solutions deliver networking and security entirely in software, abstracted from the underlying physical infrastructure and agnostic to any underlay network . NSX lends itself perfectly to a software-first strategy and the virtual cloud network . It provides pervasive, end-to-end connectivity for your apps and data, wherever they are.
Current and Future Infrastructure
Where is the intersection between current and future infrastructure? Today, your organization maintains a large mixture of not only different workload types but different physical infrastructure or no physical infrastructure at all (as in the case of cloud, hosted, SaaS, and more). Any solution should examine the intersection between solutions that are under hardware control with those that are controlled by software.
Data center operations are evolving in the same way that applications are being dis-aggregated and serviced through a set of APIs. The physical network fabric has been abstracted into a simple set of fabric functionality. Various hardware components can plug-and-play into this fabric, and various fabric techniques can be used, both new and old. The abstraction platform is dis-aggregated from that underlay, allowing independence.
Software tied to a hardware system is by its very definition tied to that fabric’s past. The software must continually support the legacy needs while attempting to deliver new capabilitie. This tie to the past slows innovation in the software, and thus causes a faster hardware churn for new service needs. This is happening all too frequently in today’s data centers.
A Changing Mindset
Ultimately, networking and security must attain the characteristics of cloud: agility, speed, manageability, and portability. Networking is no longer about brand loyalty for the sake of history. Network teams are now in a key position to get off the hardware-centric solution treadmill and to promote change. Don’t mistake this for bias toward a virtualization-only solution. The only bias should be toward the solution that best solves business problems.