TAG Video Systems recently launched its Realtime Media Platform, a new, open-source solution that monitors, aggregates, manages and uses data-driven viewer analytics. It quickly provides users with the data required to build and operate linear media systems that meet the expectations for which they were created. TAG’s Realtime Media Platform works in realtime with live video and linear video. It was not designed to work with files. In addition to realtime content, the solution is a realtime technology, meaning that it encodes, decodes and transcodes in realtime, and it provides extreme low-latency visualization and captures data generated by sources in real time. It is specifically designed to inform technicians as they make streaming adjustments to tweak and improve the Quality of Service (QoS).

TAG introduced its software-based IP Multi-Channel Monitoring (MCM) platform in 2010. It is the foundation for what TAG’s clients have been asking for - a Media Control System (MCS) for data aggregation and management. The MCS has many new capabilities and it supports SRT, NDI, Dolby ATMOS, Eluvio blockchain-based OTT, JPEG-XS mezzanine compression, AWS Elemental CDI cloud transport and new AWS instances as they become available. It also supports the PCIe-4 standard, which doubles internal bandwidth and significantly increases capacity.

At a recent webinar introducing the Realtime Media Platform, Paul Briscoe, TAG Chief Architect explained, “The MCM can receive virtually anything that comes in over IP. It doesn’t have SDI or HDMI inputs or outputs. Everything comes in IP and goes out IP. It fully supports uncompressed standards such as SMPTE 2110 and 2022-6 for live production. It also supports production on mezzanine formats such as JPEG-XS and NDI. We support all the streaming distribution formats as well as the OTT formats. This includes both their transport, the codecs they use, and in the case of OTT, a number of decryptions as well.”

Chief Architect Briscoe continued, “We also support reception using transport redundancy with SMPTE 2022-7 Seamless Protection Switching. All the monitoring we do, and all the probing of signals that we do is all done live in realtime. Nothing is deferred, nothing is after-the-fact, it is all instant realtime. All the inputs are probed, and we monitor through the probing technology to see what is going on. We look at every layer of the signal, because in IP signals are wrapped like an onion and we must unwrap that onion to see all the layers. We probe all these parameters live, in realtime while they are coming in.”

Briscoe revealed how it works. “We start by monitoring the IP transport and report any issues. We look at format-related transport because very often you might have something else wrapped up in IP, and within that are media containers. We’re probing all that as well. When we finally unwrap everything, we get down to decoding and perhaps decrypting the video and audio. Once normalized to video and audio inside the system, we can analyze it and look for Quality of Experience (QoE) metrics such as audio meters, frozen frames and black.”

Alarming and Notification
“Everything’s got metadata. There are two kinds of metadata, the kind that comes with the program content and the kind about the actual transport, and we’re monitoring it all together. Monitoring all metadata in realtime generates a huge amount of data that we deliver for alarming and monitoring. In that part of the product, the customer can decide what is an alarm and what is just an event. We provide templated monitoring where customers can create a custom template that defines what alarms are important to them. We have categorization based on severity and several notification methods when alarms do occur. We can send them by SNMP, syslog, Amazon SNS to your network, or by email. It can also display notifications on its Multiview screen in Master Control,” Briscoe said.

The Multiviewer provides realtime visualization and is historically the core of the product. It displays full frame-rate images of all live sources across all formats, SD through UHD, and it supports ABRs of various arbitrary resolutions in OTT. Multiviewer outputs can be supplied in simultaneous multi-formats, such as uncompressed for use in local control rooms, as a transport stream for distribution over IP, or as HLS for content coming in over the internet.

Exposing the Data
The Media Control System is the gateway into the entire TAG system network. MCS provides a single point of user management and access. It is also a single point of orchestration and automation control for the entire TAG system.

The MCS is bringing all the rich data from all the MCM devices, in real time, into the MCS which curates the data. It aggregates all the system logs, all the live metadata, and system statistics. It exposes that data through a ‘big data’ interface, using popular 3^rd party, open-source analytics and visualizations tools such as Prometheus, Grafana and kibana to open up the data and make it useful for customers. These 3^rd party tools are commonly used by IP professionals across many different systems, and the talent pool familiar with them is large. Users can build tools on top of the metadata for dashboard visualization, and it can be delivered to ML and AI applications for forensic analysis, audits, and forecasting.

TAG Director of Corporate Strategy Peter Wharton said, “The first thing on the customer experience improvement is to not have problems. Being able to find problems quickly through Root Cause Analysis, and therefore improving Time to Recovery, is critical. The same data can be used for SLA verification and performance analytics.”

Problem Prediction
Wharton continued, “Then we can visualize trends. This is where it really becomes valuable. Now you can predict problems before they impact the customer experience. You can see, for example, that certain kinds of errors are popping up more regularly and getting worse. Perhaps the errors are not yet at the threshold where they impact the viewer, but we know it’s getting to the point that if we don’t address them soon, that could happen. Because we have all this rich data and because many customers are now running in the cloud, it is possible to use AI and ML to drive and parse the data to be able to look inside it and do things like automated failover, dynamic scaling and failure prevention.”

“Media companies want to increase the number of customers, increase consumption, and end up with increased loyalty from those customers,” he said. “The challenge to media companies is to understand their customers and the customer experience. “What we see is companies that are using data-driven viewer analytics to drive decision making. This is a key shift in our industry. The decision-making process on the business and programming sides are both being driven directly by data, sometimes in real time.”

“If you are looking at the analytics to see if viewers stay tuned to a show after the first 10 minutes or not, you don’t know if your data is accurate,” Wharton said. The problem is that viewership data accuracy can be affected by technical content and delivery issues. “You can improve data accuracy by monitoring and correcting for data errors caused by QoS, QoE, and delivery issues,” he said. “We need to be able to measure all this data and be able to subtract it from our analytics, so the viewership analytics are accurate and track the true consumer experience and what they think of your content. This is where the TAG MCM comes in. It does the deep probing that allows us to know if there are problems across the entire delivery system.” The TAG MCS is a centralized management and data store enabling error correlation and correction of viewer analytics. The open-source analytics correlate and visualize data to correct errors faster.