Understanding Recommendations
Recommendations apply to all application traffic originating at a particular site, for a specific application category. For example, if you have 6 application categories, you might see up to 6 separate recommendations per site, depending on whether the potential exists for improvement within each category.
Recommendations are a signal that you have room to improve overall network performance, but it’s up to you to decide whether and how to take action. Here are some guidelines to inform your choices.
If WAN Insights sees a problem but doesn’t find a better path to recommend, or quality is already close to 100%, there’s no recommendation because there’s no way to improve the situation unless you increase network capacity.
Recommendations are site-level, but each site might contain potentially a large number of paths to other sites. It could be that only a few of these paths are causing most of the problems. Review the detail screens for each endpoint pair to investigate further.
The recommendations are only for outgoing traffic from that site. The site on the other end might have a different recommendation, and you will have to weigh other factors such as the number of users at each site.
How Often Do Recommendations Change?
WAN Insights is focused on long-term improvements, rather than momentary disruptions. Although the detail screens show hourly breakouts, you won’t see a new recommendation unless quality issues persist over several days, and the alternative paths are consistently better.
WAN Insights re-evaluates all sites and updates their recommendation status once per day, based on a rolling time window:
For each active recommendation, is the recommendation still good, i.e. is it still “best path”? If it is, leave it up. If not, mark that recommendation as terminated.
For sites/applications with no active recommendation, re-examine the latest 7 days of historical data. If a better path is found, generate a new recommendation. If no better path is found, no recommendation is generated.
Note that recommendations are generated by site, but there are separate recommendations for each application category. So if you have 5 application categories, and sub-optimal pathing for all of them, one site can have as many as 5 separate recommendations.
Path Quality Scores and What They Mean
Current path quality shows the network service quality that your users are experiencing now. Recommended path quality or Projected quality shows the maximum quality that could potentially be realized, using alternative network paths that are already available. Both default and recommended path quality rely on the same underlying network measures.
Path quality measures how close the Quality of Service (QoS) score is to the SLA (quality thresholds) configured for each application category. A lower path quality means that you can expect to see more SLA violations. Thus, WAN Insights suggests actions that you can take to avoid these violations.
Note that for recommendations, we deal with Quality of Experience (QoE) scores rather than Quality of Service (QoS) scores. QoE scores are averages over long time periods (7 days) and over multiple tunnels, and are derived from the QoS scores. Since quality often fluctuates strongly over time and across tunnels, a long-term average QoE score can be interpreted roughly as the fraction of time that you are within the SLA.
Load Balancing Assumptions
When recommending a set of paths, the WAN Insights recommendation feature assumes an equal load balancing on all circuits. This applies both to current path quality, and recommended or projected path quality.
Bandwidth-Aware Recommendations
WAN Insights considers both available bandwidth and existing traffic volumes when generating recommendations for path changes. Bandwidth values can be entered in vManage and have it propagate into WAN Insights, or you can upload it from WAN Insights as described on Enter or Upload Bandwidth Data.
The capacity check looks at peak traffic, so even when hourly traffic is at its highest, the recommendation is still applicable. Peak traffic refers to the traffic observed during the 10-minute interval which showed the highest bandwidth utilization over the past 7 days.
WAN Insights computes the utilization values by taking the maximum for both these values, over the last 7 days:
Existing utilization of the recommended set of paths
Utilization for the "to be moved" traffic
If bandwidth information is present, the recommendation will say Capacity Check: Passed. WAN Insights will not generate a recommendation that would, if followed, exceed this available bandwidth. In other words if a path change recommendation would result in moving too much traffic to a new interface that can’t handle it, the recommendation is not created. Where Does Bandwidth Come From? explains the different ways that capacity data can be added to WAN Insights.
If the bandwidth data is not present, the recommendations will say Capacity Unknown. The user will be prompted with a link to update capacity data directly in the WAN Insights Capacity Planning Settings page as described on Enter or Upload Bandwidth Data. After bandwidth information is supplied, any recommendation that would exceed this bandwidth is withdrawn.
Recommendations always support the bandwidth available, if known. Therefore, there is no “Capacity Check Failed” message.
Recommendation Life Cycle
WAN Insights re-evaluates site-level recommendations daily, based on the last 7 days’ worth of network data.
Taking Action on Recommendations
For the initial GA release of WAN Insights, the recommended path changes are made outside of WAN Insights, by network teams who create a localized SD-WAN policy. In other words, WAN Insights suggests what to do, and the required action is to re-configure your network at the equipment level.
To take action, you must:
Determine independently if the recommended path can actually handle all the application traffic.
Initiate the process to make network changes according to your own IT policies and processes.
Additional notes:
If there’s no recommendation for a particular site, it means WAN Insights didn’t find a better path available.
If you’re facing more immediate trouble, such as users reporting problems, use the Site Details Screen to locate the sites and see quality measures without needing a recommendation first. See WAN Insights Introductory Tour, Part 3 for guidance on responsive troubleshooting.
Which Recommendations Should I Work On First?
General guidelines for prioritizing recommendations
Optimize paths for spoke sites first, and then review recommendations for hub sites.
Look for recommendations with the greatest impact, which is calculated as the number of users multiplied by the percentage of quality improvement.
Last updated
