Raising the Quality of Experience Bar

October 13, 2021


Per Wikipedia¹, quality of experience (QoE) is defined as: “The degree of delight or annoyance of the user of an application or service. It results from the fulfillment of his or her expectations with respect to the utility and/or enjoyment of the application or service in the light of the user’s personality and current state.

In other words, a service’s experience is, of course, subjective but also dependent on and benchmarked against a person’s expectations based on past experiences.

In the 1960s, witnessing a TV broadcast in color for the first time was such a transformative experience that if people had been surveyed, they would have given the experience the highest QoE score.

Today’s viewers are accustomed to an excellent TV experience, which has raised the bar. Further improving QoE requires a relentless focus on ensuring that every building block of the video service contributes to this goal.

As TV service providers embrace new technologies to make their services available on more screens in and outside the home, they need to go beyond ensuring that the quality of experience is on-par with traditional TV services that are delivered via a managed IPTV or QAM STB. Their service needs to match or exceed what new entrants like OTT service providers offer.

Consequently, TV service providers and broadcasters are on a journey to transform their technology stack and operations. The goal is to become more nimble and agile while at the same time meeting increasing viewer expectations.

We at Broadpeak are especially focused on supporting our customers in this journey by providing products and solutions that deliver the best possible quality of experience.

In this blog we will discuss some of the prime features and capabilities of our products and solutions that reinforce QoE.

Video Delivery

The most radical transformation happening is the adoption of adaptive bitrate (ABR) streaming by TV service providers. This is being driven by the need to deliver TV services — linear TV and video on demand (VOD) — to consumer devices.

Initially ABR streaming was aimed at so-called second screens, such as mobile and tablet devices, in and outside the home, but now it includes the Connected TVs inside homes.

The shift of delivering video to the main screen using ABR protocols such as MPEG-DASH and Apple HLS has further stressed the importance of a scalable and optimized audio and video delivery pipeline for these new delivery protocols.

To enable the delivery of ABR streams in a scalable manner, service providers must leverage content delivery networks (CDNs).

The Unicast CDN

A unicast CDN is a distributed network of interconnected caches that automatically stores parts of content closer to the subscribers based on its popularity. This ensures that subsequent requests for the same piece of content can be served from a cache rather than having to pull the content all the way from the source/origin.

The base concept of a unicast CDN is that of a reverse proxy. Each cache is a reverse proxy that terminates a request from a playback device and attempts to serve the request with local cached instances of the content in the first place. In case the requested content is not present locally it attempts to retrieve the content from other caches or from the origin. The retrieved content is cached along the retrieval path so that subsequent requests can be served locally.

This is a dynamic method of caching that is driven by the popularity of the content and by the size of the different caches. Indeed, as the cache fills up the oldest and least recently requested cached content is removed from the cache.

When implementing a CDN for ABR streaming, it is important to place the CDN caches at appropriate locations in the network based on the expected peak playback requests originating in the different parts of the network. The size of these caches should be based on the expected popularity of the content.

Hierarchical Cache Topology

Given the dynamic nature of the caching it is important that each playback request is directed to the most appropriate cache. Additionally, it is essential to build a hierarchy of caches to optimally distribute the content across the network according to its popularity. Here, having a CDN topology and functionality that can serve the most popular content from the edge of the CDN and maintain a good portion of the long-tail content in more intermediate locations is essential to achieve optimal trade-off between storage/processing and network bandwidth costs.


Caching origin topology hierarchical
Figure 1. Hierarchical caching topology illustrated

Also depending on the cache versus egress ratio, different types of cache servers and quantity of servers can be used in such a topology to build a cost-effective CDN.

Broadpeak’s advanced CDN has an elaborate request routing mechanism that also leverages knowledge of the hierarchical cache topology and network connectivity between the caches.

This is enabled by defining a CDN topology model where:

  • Individual cache servers are grouped in logical cache elements
  • Specific backup links between groups of cache servers are defined to fail over between groups of server

This allows for the optimal use of the CDN and network resources to deliver a video service with the highest possible QoE.

Session-based CDN

Besides implementing an optimal topology of caches, it is also important to utilize these caches as efficiently as possible. Given the dynamic nature of the caching, you can achieve maximal utilization by ensuring that as much as possible requests for the same piece of content are mapped onto the same cache server. Broadpeak’s advanced CDN creates and maintains a session context, which is used in subsequent HTTP requests and is used to map onto the same cache instance. Additionally, this session context identifier can be used to generate detailed session playback reports and statistics.

Video cache servers

Besides ensuring that sessions are consistently mapped onto the same cache server Broadpeak’s advanced CDN supports mechanisms to selectively associate sessions to specific types of cache servers according to the type of content that needs to be served. Indeed, a request for a VOD asset is best served from a CDN cache with a large cache. However, a request for a linear TV channel segment is better served by a CDN cache with high throughput and only minimal caching capability.

If you know that some assets will be popular from the start, then having a CDN that enables you to pre-publish/cache assets in select CDN caches can help drive the overall efficiency of the CDN. With this approach, you can avoid unnecessary cache pollution and consequently realize faster access to the content. The throughput of an individual cache server can be further optimized by implementing an optimal cache tiering in the server itself too. Here you must ensure that the cached content is distributed dynamically across the different storage subsystems present in the cache server, such as the NAS, DAS/SAS spindles, NVMe, and RAM.

All of the optimizations offered by Broadpeak’s advanced CDN allow you to achieve Time-to-First Frame (TTFF) for VOD and a short zap time for linear TV. This is critical to delivering a high QoE.

Bandwidth Management

When deploying a unicast CDN it is important for service providers to remain in control of the content delivery infrastructure, especially in cases where there is unforeseen high demand. However, with ABR streaming it is in fact the video client/player that independently decides on the bitrate to request from the CDN. Consequently, we end up with different independent clients requesting content and competing for the same bandwidth. The bandwidth bottleneck can reside in the home, but at some peak moments the CDN caches themselves might get overloaded with requests. In such an event all clients downstream from the bottleneck will get switched back automatically to a lower quality bitrate, independent from the type of device and/or subscription. In other words, the quality of experience is negatively affected for all subscribers.

Hence, a CDN that enables service providers to selectively control which devices and subscribers get access and priority over others is quite essential in offering a premium TV service.

Broadpeak’s advanced CDN offers several bandwidth management capabilities to achieve this.

  • Set Type of Service (ToS)

By defining different categories of service/subscribers and having the packager/cache set the IP TOS value in the IP header accordingly, you can ensure that if there is congestion on the network the traffic with the higher TOS value gets precedence over the lower one. In essence, you can effectively differentiate between the services/customer categories.

  • Rate shaping

As the number of TOS values are limited, you might still end up having to share the same TOS category with other IP traffic. To ensure stable video delivery, Broadpeak supports rate shaping on our advanced CDN. This ensures that the aggregate bandwidth for a given TOS value never exceeds a set value.

  • Adapt the bandwidth based on the context

With Broadpeak’s Server-Side Segment Selection Streaming (S4Streaming) feature on our advanced CDN, we enable more fine-grained control of the bandwidth on an individual session basis.

As such, we can assess the overall bandwidth consumption of a CDN cache and ensure that sessions fitting certain criteria are offered a higher bitrate then others. This is all done in a dynamic manner, even for ongoing sessions. As such, it becomes possible to ensure, for example, that at all times the main screens have access to streams in the highest quality/bitrate, but that the secondary screens only get access to those higher quality streams when there is sufficient aggregate bandwidth available.

As we leverage an aggregate view of the bandwidth, we can also better allocate the bandwidth to the individual sessions and consequently drive a better utilization of the caches and links while maintaining the highest possible quality of experience appropriate for the device.

Equally, within a single ABR session it is difficult — especially in the case of Low-Latency ABR streaming (LL-DASH/LL-HLS) — for video players to correctly estimate the available bandwidth on the path when deciding which video quality to request next.

With the S4Streaming feature we implement the bandwidth estimation algorithm and the selection of the next segment to send/request on the server side (CDN cache).

Besides control over the individual sessions, we also support the use of geo-blocking and the capability to limit the number of simultaneous stream sessions per account.

The Multicast CDN

With the shift toward ABR-based delivery also to the main screen, service providers are having to deal with scaling their CDN to handle the increased number of playback sessions as well as the higher bitrates required to ensure the best experience on big-screen TVs.

Although the death of linear TV has been predicted over and over, the reality of the matter is that it still represents most of the prime-time viewing. This is especially true for sports events. For regular linear TV programs there is quite a predictable pattern for the playback concurrency. Consequently, sizing a unicast CDN can be done reasonably correctly to ensure that every day of the week the linear TV service can be enjoyed by the subscribers. However, true live events and especially sport broadcasts have a less predictable traffic pattern and make it difficult, if not impossible, to size the CDN upfront.

Across the industry there is a lot of talk about so-called elastic CDNs, where additional CDN caches can be enabled in the network to handle the peaks in demand.

However, this assumes that at those moment of peaks the underlaying hardware platform with the right characteristics is available to instantiate the CDN cache on and that the cache can be filled sufficiently quickly to absorb the playback peak.

Given that most TV service providers do not have such spare capacity available they have shied away from adopting ABR for delivery of linear TV to the main screen and have continued to leverage their legacy IPTV multicast or QAM broadcast technology.

However, with the advent of multicast ABR (mABR) you can scale linear TV by means of IP multicast, using ABR protocols to deliver the service to retail devices.

Indeed, in a multicast ABR scenario, there is a central function in the network, called the transcaster, which retrieves the ABR protocol constructs (i.e., manifest and segments) from a regular unicast origin/packager. The segments/chunks are transmitted on a multicast group/channel per rendition/quality as they become available.

CDN Edge 5G. At the subscriber home, either a gateway or STB with a mABR agent is deployed.
Figure 2. High-level Multicast ABR Architecture.

The mABR transcaster also ensures that the manifest is transformed/augmented in such a way that when this manifest gets loaded by a client device it will send the segment requests to the local mABR agent rather than the origin. This agent knows from the request parameters which multicast group/channel to use to retrieve the appropriate segment. In other words, the mABR agent acts as a local cache of the central origin to make the content available to all devices in the home.

Broadpeak has been at the forefront of mABR development and deployment with its nanoCDN solution. nanoCDN not only enables delivery of linear TV channels in ABR but also retains the same content protection and enables the same targeted advertisement capabilities of a pure unicast-based delivery.

By combining Broadpeak’s advanced CDN with the nanoCDN, you can deliver both on-demand and linear TV services in a scalable and cost-effective manner to the widest range of retail devices. With this solution, you are improving the experience for subscribers as they can now use their device and platform of choice.

Low-Latency Service Delivery

We already touched upon the importance of being able to deliver video swiftly, so the playback can start as quickly as possible. Beyond the desire to reduce the time to the first video frame there is also an overall importance to reduce the latency. Due to the specific way video segments are constructed and how video players process them in ABR we quickly end up with latencies of 30 or more seconds. Clearly for on-demand this is not so critical and even for most linear TV channels these latencies are quite acceptable. However, as service providers are increasingly delivering live TV, especially sports using ABR, the latency becomes somewhat of a problem for the quality of experience. Imagine hearing your neighbors cheer for a goal and not actually seeing it on the TV until 30 seconds later!

Consequently, the industry has been exploring ways to reduce the latency and witnessed the advent of initially a low-latency version of MPEG-DASH, commonly referred to as LL-DASH or CMAF-CTE. Subsequently, Apple evolved its HLS specification to enable lower latencies by introducing LL-HLS. Both low-latency protocols help reduce latency down to practically two to three seconds.

The adoption of low-latency protocols, however, also requires the appropriate support by the CDN. The origin/packager needs to support the new ABR protocol specification in terms of the manifest and the creation of the chunks/segments. Additionally, the core principle of low-latency streaming is no longer based on waiting for the full segment of several seconds to be available at the origin, but here smaller sub-second chunks are sent from the moment they become available from the encoder. Thus, CDN caches need to ensure timely delivery of these chunks too.

In cases where the operator wants to scale the delivery of linear TV by leveraging multicast ABR it is essential that the mABR solution also supports these new low-latency protocols to ensure the best possible customer experience.

Broadpeak’s relentless focus on innovation has resulted in an advanced CDN that is low-latency ready and a nanoCDN solution that supports both LL-DASH and LL-HLS.

Operators who choose Broadpeak’s CDN solution can offer their subscribers the industry’s best quality of experience.

Third-Party Content

TV operators are increasingly complementing their TV service offering with OTT content from third-party providers by integrating OTT content in the main TV experience. Consequently, it has become important for TV service providers to offer third-party content with the best possible QoE. Moreover, there is an incentive for OTT providers to be able to deliver the best possible experience to TV service provider customers.

TV Providers who adopt CDNs with Streaming Video Alliance (SVA) Open Caching support can quickly provision new 3rd party content providers and ensure that playback requests for this 3rd party content is carried by the TV service provider’s CDN. Equally, these 3rd party content providers are given access to standardized performance measurement data like with a global CDN. This represents a win-win for both the TV Service provider and the 3rd party content provider. Additionally, the subscriber is now able to experiencing a superb Quality of Experience for OTT content too.

Third-Party Content CDN Mabr CACHE caching SVA ORIGIN
Figure 3. Third-Party content onboarding using SVA Open Caching.


Broadpeak has been at the forefront of the development of Streaming Video Alliance (SVA) Open Caching specification. The company performed the world’s first live demonstration of SVA Open Caching APIs at the Streaming Video Alliance’s first 2021 quarterly meeting. The demo was powered by a complete setup that included the video content source (an origin packager provided by a major content provider) and Broadpeak’s umbrellaCDN™ CDN selector on the client side. Broadpeak’s advanced CDN consisting of the BkM100 for CDN management and BkS400 local cache server were used during the demo on the service provider side.

With the advent of OTT content providers specializing in live/linear TV, we are also witnessing an increasing desire by TV service providers to integrate live OTT services into the user experience. Thus, Broadpeak is expanding its multicast ABR solution also, to enable the smooth onboarding of linear OTT services and to ensure that peak events like the NFL Super Bowl or the UEFA European Football Championship can be delivered in a scalable manner with the best possible QoE for the enjoyment of subscribers.


Although QoE is a subjective concept it is important for TV service providers to continue to focus on delivering outstanding quality, especially in these turbulent times where many new alternative avenues for content are available. For TV service providers that operate their own network infrastructure it is essential that they can leverage this asset to the maximum extent possible. Using a CDN solution optimized for the delivery of video using ABR streaming is key.

Broadpeak offers the ultimate CDN technology, providing optimized unicast delivery with our advanced CDN and enabling unlimited scalability of linear TV via IP multicast with our nanoCDN solution.

The experience is further enhanced by adopting low-latency streaming protocol support, advanced bandwidth control, and last but not least by supporting open interfaces that allow the smooth onboarding of third-party content providers.

¹ Source

About The Author

Pieter Liefooghe is a Business Development Director at Broadpeak. Before joining Broadpeak Pieter held the SVP Product and Operations position at THEO Technology, a video playback component vendor. During his three years at Ericsson/Mediakind he was in different roles, notably Media Delivery Solution Area CTO covering primarily their Cloud DVR solution. In the 11 years prior, at Alcatel-Lucent/Nokia/Velocix he was active in different organizations and roles, lastly as the Product Management Director responsible for their Cloud DVR solutions. Prior to joining the corporate world, he was active for over 10 years in academia doing applied research in the field of telecommunications and streaming media in particular. Pieter holds a PhD in Applied Sciences from the Vrije Universiteit Brussel, Belgium and an Engineering degree from the same university.

Do you want to discuss with us your QoE challenges?