DX NetOps Intent Based Networking for SD-WAN

by May 20, 2019

The complexity of Software-defined Networks (SDN) and the rate of change executed by controllers require a way to provision based on what a network should do and how it should behave. This approach is called Intent-based networking (IBN) and is fast replacing manual or even script driven device and network deployment and configuration. The ability to apply an intent for how a network should behave is the key to SDN and allows Enterprises to maximize performance balanced with cost. The most popular IBN based technology today is SD-WAN and in this blog we’ll look at how DX NetOps makes sophisticated IBN not only possible, but easy.

Network Engineers and Administrators intend to configure a network to act a certain way when delivering traffic. This is known as traffic-shaping and has been around for a long time in the form of QoS and similar technologies. Network Administrators have used QoS to “intend to” deliver optimal network resources for critical applications and services. Through QoS policies, differing traffic classes can take advantage of network resources at a higher priority than less-critical traffic. Less-critical traffic can be dropped when resource contention is tight.

This ensures the high priority traffic is not impacted while crossing congested network links, but what about that lower priority traffic that’s dropped? What type of overall user-experience of the network is being experienced and at what cost? How can these policies adapt to change in the network automatically?

While QoS has provided some level of intelligent “performance-based routing”, SD-WAN technologies have taken this a step further. Network Engineers can now define networks with more sophisticated intentions than dropping lower priority traffic for higher priority traffic using static policies. Now SD-WAN can be configured to optimize the flow of application traffic over the WAN; not just based on performance, but optimized to take advantage of lower cost technologies (such as broadband over MPLS).

Why is this so important? Because the availability and reduced costs of high-quality transports other than MPLS creates a strong opportunity to minimize costs while ensuring a superior customer experience. These cheaper connections however lack the guarantee of SLA’s from technologies such as MPLS and therefore it’s harder to predict how performance will behave. And beyond predictions, there is often significant variability in Broadband quality based on time of day and day of the week. This makes it difficult to take set policy thresholds that can adapt to normal behavior over time.

Take a look at this graph measuring latency of a broadband connection over the course of a week.

You’ll notice the variability of minimum and maximum values based on day and time. If I wanted to ensure my Broadband connection was utilized as often as possible, you might think a simple static threshold of 200 msec would ensure good user experience during business hours with off-hours switching to an alternative to handle the peaks.

But what about ensuring my Broadband continues to be used during the weekend as well? I could assume my performance will be consistent and therefore increase the latency threshold to 300 msec, intending that it should behave better during the week. If latency of 290 msec during business hours impacts user experience, how can I have the best of both worlds?

DX NetOps and the DX AIOps solution from Broadcom can help. In our example below, using intelligent baselines that compare current performance to same day of the week and hour of day, I can leverage this information to better understand if the behavior is predictable and if my intent to rely on good performance based on historical norms will make me feel good about my 300 msec threshold.

This still presents a problem however in situations such as the one below. How can I ensure that during the week I take advantage of other transports to meet user experience goals while maximizing my use of broadband during non-business hours.?

Using the intelligent analytics within DX AIOps for Networks, I can use a dynamic threshold that focus on deviation from baseline to trigger and alarm when my performance is outside normal ranges. These alarms can then leverage the automation APIs in Operational Intelligence from Broadcom to drive policy changes to the controller.

Through baseline analytics and automation APIs, DX NetOps can provide a more intelligent, dynamic SD-WAN policy threshold that is more adaptive to the changing conditions of today’s network and applications.

Check out our latest webcast on SD-WAN Analytics:

DX NetOps: SD-WAN for Minimized Costs and Reliable App Delivery