7 steps to SD-Wan Deployment

Networking architecture is now seen as a linchpin of digital transformation. Deploying SD-WAN architecture may bring cost savings, performance benefits and ease of management.

Most enterprises are now undergoing massive shifts in the technology they use and the processes they adopt.

That’s because, traditionally, most companies have used manual and analog processes. But technologies like cloud and edge computing, mobility, artificial intelligence and Internet of Things-enabled devices have upended these long-standing processes.

Digital transformation captures the massive shifts that enterprises are undergoing as they digitize and automate operations. Digital transformation integrates technology to solve traditional business problems with automation, digitized processes and AI. Digital transformation can streamline how businesses operate while also building competitive advantage and business differentiation.

One of the first steps to this automation is virtualizing IT infrastructure, particularly with network virtualization, according to Gartner Research on the role of network virtualization in digital transformation. For most enterprise IT departments, the easiest place to start virtualizing a network is the wide area network (WAN) edge. Software-defined WANs bring relatively swift cost savings and performance benefits. Virtualization eliminates hardware, enables more flexible management and more.

If your business is ready to adopt software-defined wide area network (SD-WAN) technologies, the next logical question is “How?”

No two SD-WAN architectures will be designed, built and managed identically. Multiple factors will alter deployment decisions, including the makeup of a remote site, enduser requirements and budget limitations. But here are seven steps in deploying SD-WAN that can help you ask the right questions. Those steps are the following:

1. Calculate the maximum number of remote sites
2. Right-size the deployment based on number of users and anticipated WAN usage
3. Analyze applications/services/workloads
4. Determine which WAN connectivity options are available
5. Choose a deployment model
6. Plan for post-deployment WAN data-flow collection
7. Continuous improvement based on analytics

Let’s look at each of these steps in more detail.

1. Calculate the maximum number of remote sites. The number of viable SD-WAN deployment options increases in conjunction with the number of remote sites an organization has. Smaller SD-WAN use cases tend to keep the overall architecture straightforward by using out-of-the-box configurations. For simplicity purposes, these types of deployments also typically use a star topology that brings all traffic back to a central location. Most organizations opt to go light on design complexity because they lack the budget, manpower or need for a more robust design.

When it comes to how intricate they can become, large deployments feature greater flexibility. Much of this depends on the criticality of the remote sites in terms of business continuity. The more critical remote locations become to a business’s bottom line, the more sophisticated an SD-WAN deployment should get.

2. Right-size the deployment based on number of users and anticipated WAN usage. The number of end-users—as well as where their applications and data reside—both play an important role in determining how to design and size an SD-WAN. For example, if a remote site has hundreds or thousands of users, but applications and data are managed on-site, WAN intelligence and optimization may not be as important compared with a 10-user site that uses mission-critical apps located farther away from a WAN. So, while from a throughput perspective, head count is important, be sure to gauge per-user need and business criticality of a WAN.

3.  Analyze applications/services/workloads. Next, evaluate the number, types and importance of all anticipated applications, services and workloads that will traverse the WAN. Conduct research to identify each application, how this app interacts with distributed resources across the WAN and the minimum network requirements. Real-time streaming protocols such as voice, video and access to high-performance databases will require these application data flows to be identified, marked and prioritized over a WAN.

The make-or-break secret to a successful SD-WAN deployment comes from this application analysis. Only through this analysis can IT departments determine whether apps might drain SD-WAN resources or require faster speeds; only by understanding these application requirements can you properly set up your SD-WAN. Without performing a thorough audit and prioritizing applications according to importance and requirements, intelligence of an SD-WAN deployment lacks the information to make proper routing decisions.

4. Determine which WAN connectivity options are available. Once you calculate throughput and latency requirements based on the information gathered in the three previous steps, investigate which WAN connection options are available at each location. Don’t forget: SD-WAN technologies must have two or more WAN links for SD-WAN AI to have a choice of paths to use. If your branch offices are all located in modern, urban areas, the number of private WAN and/or broadband carriers may be plentiful. In this situation, choosing WAN connectivity types boils down to answering the following questions:

• What throughput, latency and resiliency is needed today?
• What options best allow for the type of scaling up and/or down anticipated?
• Which options best fit the budget?

If, however, your remote sites are in rural areas, you must also address the constraint of limited WAN options. No stone should be unturned in this situation. Fortunately, the intelligence underlying SD-WANs can still use low-bandwidth and higher-latency links–while squeezing the most out of them. Thus, options that are often initially discounted (such as LTE and satellite broadband) could end up being viable connectivity options in some locations.

5. Choose a deployment model. All prior steps in this process deal with the framework in which SD-WAN will operate. When that is complete, the fifth step is to explore SD-WAN deployment options and which model best fits. There are three basic deployment models to choose from.

First, in-house IT staff could negotiate pricing and enter into WAN connection contracts directly with the vendor. Once complete, an SD-WAN could be chosen, deployed and managed fully by the company’s IT department. Alternatively, many IT decision makers have chosen to let an SD-WAN managed service provider (MSP) handle all aspects of the WAN, including the relationships with WAN circuit providers, SD-WAN deployment and all ongoing maintenance.

Lastly, it’s increasingly popular to deploy a hybrid solution that splits duties between in-house IT staff and an MSP. In this scenario, an MSP manages the underlying WAN infrastructure by monitoring performance of the WAN connections and preemptively opening trouble tickets with WAN circuit carriers on the client’s behalf. Then, the overall WAN policy creation, management and security posture is left to in-house IT staff, who better understand users’ needs. As Gartner has noted, many IT departments are now responsible for managing service delivery, not for managing traditional IT infrastructure.

6.  Plan for post-deployment WAN data-flow collection. It’s important to note that even after deploying SD-WAN architecture, the technology requires ongoing maintenance to operate efficiently. While artificial intelligence within an SD-WAN platform eliminates numerous manual processes, it still must be fed pertinent information regarding changes in user needs, growth or business demands. This information needs to be collected and curated so that the intelligence built into SD-WAN architecture can understand the changes and make necessary adjustments in data flow policy.

7.  Continuous improvement based on analytics. When an IT department deploys an SD-WAN architecture, it usually behaves according to the information originally provided to it. It’s up to a network administrator to determine when new policy information (based on business needs) should be introduced that will alter an SD-WAN’s behavior. The data collected should be analyzed and then re-introduced to the SD-WAN platform at regularly scheduled intervals.

Changes to connectivity type must also be re-evaluated on a set timeline. New WAN connection options that improve WAN end-users’ experience may become available. Additionally, the number of users, application types and criticality of apps/data will likely change over time, increasing or decreasing the amount of throughput, speed and jitter at each location. It’s important that continuous improvement is planned and appropriately scheduled to ensure the intelligence of an SD-WAN to is effective.

Deploying an SD-WAN

As you consider deploying an SD-WAN architecture, don’t lose sight of how the technology will be used in your particular environment. While each step outlined doesn’t need to be accomplished sequentially, the steps can be broken into pre-planning, architecture and ongoing management phases.

During pre-planning, consider the connectivity types of each branch. Also, calculate the number of endusers and their unique application requirements. The architecture phase includes factoring geographic location, user requirement and deployment model to best achieve your goals.

Lastly, in the ongoing management phase, you need data collection and analytics to optimize your SD-WAN architecture on a continuing basis. Then, based on the analytics results, rework network policy to fit business needs. If these steps can be accomplished in the phases outlined, you’ve set yourself on the right track for an enterprise-grade SD-WAN that will perform for years to come.