Across industries, everyone is talking about AI-based solutions and how they can improve their business. This includes streaming service providers, who are enticed by their potential to solve technical, resource, and cost-based challenges.
Of course, not all AI use cases will pan out, but utilising it to improve specific aspects of business operations is certainly on the horizon. In the context of streaming, that’s exactly the case for OTT monitoring automation.
Today’s monitoring obstacles call for OTT monitoring automation
One of the biggest challenges in monitoring is time. Issues need to be identified and resolved quickly, ideally before viewers even experience them. To do this with large streaming video data sets requires both skilled operations engineers, who identify connections between data points and where to drill deeper, and a lot of eyes. No single engineer can see all the potential issues. OTT monitoring then makes significant demands on available resources within the NOC.
So, what if AI could carry out some of the initial work of identifying the relationships between data sets, bubbling up potential issues (based on historic comparison with known problems), and determining what data is necessary for ensuring a high QoE?
Data is the solution, and challenge, to streaming video QoE
The streaming tech stack is a veritable treasure trove of data. Every component, from virtualised encoding instances to CDNs, provides rich data about what’s happening within that component. This data is often accessible through APIs and can be dumped into vast data lakes, combined with data coming directly from the player. Against this, analytical and dashboard tools like Datadog or Looker can visualise the data for use by network operations engineers. Without all of that data, it would be near impossible to uncover and resolve issues.
However, that is also the downside. Sifting through such a high amount of data, such as drilling into a specific area of the dashboard, can be a daunting task. The challenge, then, is to have all the available data but be able to quickly identify issues. Part of that can be done within the visualisation tool using thresholds and other techniques to visually identify issues (i.e., red, yellow, green). Dealing with the raw data, however, requires a new layer in your process: OTT monitoring automation.
Putting AI to work for OTT monitoring automation
The best way to understand where AI comes in is to look at the four layers of a monitoring harness:
- Connectivity: In the first, topmost layer, data sets are connected. Ultimately, this is a decision-making process of including or excluding datasets within the harness
- Visualisation: In the second layer, those data sets are laid out linearly, representing progress through the workflow, ideally from acquisition to playback
- Integration: In the third layer, individual data sets, showing high-level information in the visualisation layer, are linked to the tools that provide more detailed information about the specific data point
- Relationship: Finally, in the last layer, data elements across data sets are related together so that the harness not only connects all the disparate data together but relates them. For example, connecting the data elements that represent a user session
This new AI-powered business automation layer is placed between connectivity and visualisation and carries out two primary tasks. The first is to automate data capture. In some cases, streaming components may not expose or even provide diagnostic data, which is critical for monitoring. Synthetic monitoring probes, deployed throughout the workflow, can automate the collection of that data. The second task is to implement business logic against the collected data. For example, rather than collecting all of the data from a specific component every second, perhaps the monitoring harness only needs every 10th record. Further business logic could be employed to standardise or calculate specific business metrics.
👉 Read more: Monitoring Harness Whitepaper
What a business automation layer means for OTT monitoring
Ultimately, OTT monitoring automation takes care of all the manual post-processing tasks once data has been received. This significantly reduces the time and effort that network operations engineers need to spend generating observability and identifying issues across the workflow. The results are game-changers:
- Enabling one-click root cause diagnosis of both software and hardware error sources so that identification and resolution of issues impacting QoE is sped up
- Reduction of alert fatigue as individual alerts are combined into more high-level issues (i.e., rather than alerting each time there is an error, aggregate errors can be alerted using visual cues)
- Identification of issues proactively based on gradual alerting (i.e., red, yellow, green)
This is also a great place to leverage AI for OTT monitoring automation. Rather than just selecting all the data, AI within the business automation player can selectively grab records based on heuristics of previous data patterns that were identified as an error. With this in place, alerting can happen even before the data ever hits the dashboard.
Moreover, AI within the business automation also works downstream: Data points can be related based upon real-time analysis, potentially bubbling up observability within the dashboard and saving network operations engineers further time trying to identify trends.
OTT monitoring automation all begins with your harness
In the most basic OTT monitoring setup, network operations engineers use a dozen different dashboards hooked to siloed data sources. Monitoring is chaotic and inefficient. A more advanced approach includes a monitoring harness that connects all the data sources into a single datapipe, which is fed into a single dashboard. This is better but doesn’t yield the best results yet.
That’s where the latest and most sophisticated monitoring system comes in, one that leverages OTT monitoring automation. It enriches the monitoring harness with a business layer, incorporating AI, ML, and other approaches. This improves the way data is captured and made available to visualisation tools, ensures alerting is managed, and leads to superior observability and monitoring efficiency. All of these combined make sure issues are addressed before they ever surface in a viewer’s streaming experience.
Of course, building a monitoring harness and enabling automation isn’t easy. That’s where Touchstream can help. By using our VirtualNOC and ISP Insights, you can get to that last scenario in no time at all.
To find out how Touchstream can help you improve and automate your operations, book a demo now.