Advanced Analytics Deliver Consistent Video Quality in the Cloud

Once requiring its own specialized networks, television is rapidly becoming just another cloud-based service being delivered by general-purpose IP networks, both wired and wireless. But streaming video, nonetheless, is still a big challenge for network operators. Sharing a network with other content and applications raises both traffic engineering and security challenges. Customers have high expectations from years of rock-solid cable TV service, so service assurance is critical. One of the key pieces in delivering customers the quality of experience (QoE) they expect is a new breed of network analytics.

In the coming five years, the networking industry will be moving from 4G LTE to 5G, the new wireless standard. Promising +10X download speeds, 5G will open up the world of mobile television. While this is getting a lot of attention in the television industry, one of the other significant shifts that will come with 5G is a move to a cloud-based architecture across both mobile and fixed networks based on virtualization and software-defined networking (SDN).

Virtualized, SDN networks have long been used in cloud data centers, but the technology is swiftly migrating to the wider area network and, with 5G, to both wired and wireless networks. One of the reasons this approach is being embraced is that SDN networks running on virtual platforms are cheaper to run, not simply because they use general purpose computing platforms instead of specialized networking hardware, but they also use the physical resources more efficiently.

This efficiency is the result of them being more dynamic. SDN makes IP networks far more flexible and scalable. They can re-allocate under-utilized hardware resources instantaneously to meet whatever needs arise. Thus, if Ninja invites Drake to play Fortnite with him, instantly bringing 500,000 new viewers to his video stream, the SDN controller can spin up the network processing resources to meet the sudden spike in demand.


Dynamically re-assigning IP resources to meet sudden shifts and spikes starts with good analytics. However, the analytics traditionally used in data centers aren’t enough. The wide area network outside the data center is a jungle made up of a patchwork of telecom operators, ISPs, MSOs, long-haul carriers and content delivery networks (CDNs). Traffic between them traverses a continually shifting constellation of peering points. Video streams can come from anywhere, and demand from viewers can shift in the blink of an eye.

Making matters even more challenging, the traditional network analytics approaches aren’t sufficient. Wide area networks have depended for decades on Simple Network Management Protocol (SNMP) to understand what was going on in the network. But as its name suggests, SNMP isn’t up to the complexity and the real-time demands of today and tomorrow’s networks.

The other key analytics technology is deep packet inspection (DPI). Deployed in routers with the processing power to look inside each packet at line speed and identify the payload, DPI can be expensive. Thus, operators tend to deploy the technology to do spot checks at key points in the network. However, its biggest challenge is that the majority of video streams are now encrypted, so DPI has no way of identifying the payload.


The good news is that there is a wealth of network data that can be used to bolster SNMP and DPI. The problem is that this data is currently collected in silos across multiple different systems—making it extremely arduous (or impossible) to use them for better planning and traffic engineering. Thankfully, recent big data techniques, such as streaming vector, column-store databases, are coming to the rescue.

Using only public available data of the same sort that Google collects, it is now possible to literally map out the entire internet and catalog the development, over time, of everything happening on it. For instance, it is possible to map IP flows from source to destination. If you know the destination, for instance, a Netflix cloud server at AWS, then you know what service your customer is accessing. With this technology, operators are able for the first time to understand what over-the-top applications are traversing their network and how they are impacting it.

With this holistic view of the entire network in real-time, operators can adjust their resources for dynamic service assurance. They can also work intelligently with partners upstream, such as content providers and cloud providers, as well as downstream, such as neighboring networks and CDNs, to adjust their own resources. Is peering point A acting as a bottleneck, while peering point B is under-utilized? Would a different cloud provider better serve a content provider, such as Netflix, in a newly served market?

As television moves into the cloud era of 5G, it is leaving the very predictable confines of its purpose-built cable networks to become part of the world wild web. While there are all kinds of good reasons to go this route, it also raises issues around network management and service assurance for increasingly demanding customers. Happily, network operators, content platforms, cloud providers and distributors have new tools to take the guesswork out of network planning and troubleshooting. These techniques are already being deployed in 90 percent of US cable operators. Along with SDN and network virtualization, which are coming with 5G, they will form part of the core technology of this exciting and challenging new world.

Kevin Macaluso is General Manager, for Nokia’s Deepfield group.