Video delivery solutions: Building the infrastructure of tomorrow

We talk a lot about the challenge of delivering quality online video. With viewer numbers skyrocketing and UHD streaming becoming increasingly standard, the stakes are greater than ever. So what’s being done about it?

Until recently, the solution was to do more unicast– i.e. to multiply peering points by installing caching servers and placing them as close as possible to ISP gateways. Today, the largest CDNs have nearly 200,000 servers placed strategically around the globe.

And that’s still not enough.

Networks may very well need to triple their capacity in the next few years, and traditional unicast delivery is proving both inefficient and insufficient for this level of traffic.

Fortunately, a range of innovative solutions has brought promise of expanding our current distribution model. They span IPTV managed networks, unmanaged public Internet (OTT) and mobile, and are the fruit of collaborations between content delivery networks, operators, and upstart tech companies working to imagine the delivery architecture of tomorrow.

As one piece of this collective effort, Streamroot would like to take a look at some of the delivery technologies in use and under development today to better understand where the market is going.

This first piece offers a broad overview of new distribution solutions and some insights into the changing landscape. In the upcoming months, we’ll invite experts in each of these areas to offer a more in depth look and their visions of what’s to come.

CDN, multi-CDN and hybrid CDN solutions

worldmap showing MULTI-CDN Location

In a nutshell

Despite its limitations, unicast CDN distribution will likely remain an important part of large-scale online broadcasting for years to come. Over time, CDNs have worked to optimize their networks of servers. DNS-based request routing, redirection and load balancing all contribute to today’s intelligent distribution.

Minimizing redundancies in the client-server relationship has come a long way. However, as further optimizations within a single network of servers become increasingly difficult, many companies are looking to innovate horizontally by combining existing CDN and distribution solutions in new ways.

Multi-CDN consists of using more than one content delivery supplier, typically for covering areas in which the main provider has fewer points of presence. The new multi-CDN network is then either managed by the broadcaster or by an aggregating provider that also manages traffic, load balancing, and analytics. Similarly, combining cloud storage solutions offered by Amazon, Google, and other providers with CDN-based delivery is also making headway as a more cost-effective alternative to a pure CDN model.

Suppliers and use cases

Multi-CDN and hybrid cloud/CDN infrastructures are relatively common for large broadcasters. Cedexis, for example, offers highly customizable multi-CDN / cloud optimization solutions. In addition to other innovative features, Cedexis allows the CDN to be switched according customer parameters: time of day, affiliate CDN / ISP arrangements, cost-based volume optimization, client device bitrate and more.

Pros and cons

Multi-CDN and hybrid CDN/cloud solutions provide better latencies and startup times on a large scale, especially where different providers offer complementary coverage. It is now common for providers to be able to switch between CDNs / clouds mid-stream, which is a significant advance for both quality of service and price.

These systems have the advantage of being relatively nonintrusive if not linked to the client-side player. On the flip side, if not linked to the player, they are also less precise, as latencies and other metrics are calculated for limited number of endpoints – which are servers and not actual client devices.

The difficulty with these technologies lies in choosing the optimal switching criteria: Should switches be based on price? Quality of service? A little of both? How accurately are these parameters actually measured? The precision component is also a tradeoff: How much data should be collected, and how much time and money can I devote to processing customer data?

Finally, these solutions remain within the realm of unicast distribution, so although they may alleviate network congestion issues, they ultimately fail to address the inefficiencies of one-to-one client-server exchanges.

Which brings us to our next point.

Multicast

Unicast, broadcast, and multicast differencies

In a nutshell

Multicast delivery has been around for decades for corporate use and in some IPTV managed networks; however, it has also recently gotten attention as a means to drastically reduce bandwidth for OTT live events.

In a multicast system, the content source transmits video content as a sequence of UDP packets to a number of receivers, each responsible for a group of users. The packet is transmitted from the source to each receiver only once; the receivers are then responsible for transmitting the content to users within their group.

When receivers are on different networks, multicast-enabled routers must use a routing protocol to build delivery trees and enable content forwarding to receivers’ groups of clients (the Internet Group Management Protocol, among others). This protocol is also responsible for keeping up-to-date information on the status of the multicast group: determining which members are still active, which content the group wishes to receive, etc. On the client end, the video player stack must be configured to support subscription to multicast groups.

Suppliers and use cases

The greatest potential for multicasting today lies within ISP networks themselves. CDNs are working to gain access to Internet service providers’ multicast capabilities to provide hybrid unicast/multicast solutions. Akamai, for example, may very well use its recent acquisition Octoshape to serve content via multicast to its edge servers, where the streams would then be reconverted back into unicast for delivery deeper in the network or directly to consumers.

Meanwhile, Broadpeak’s nanoCDN technology uses this principle to deliver multicast to users watching the same content within the same region. Optimizing the operator’s network, multicasting can replace the thousands of unicast connections that would be required to deliver to the region. The content is delivered via multicast to the home gateway, where it is then converted to unicast and can be shared among devices.

Marginal cost of an additional user: unicast vs. multicast delivery

Pros and cons

Multicast helps minimize network traffic and optimize bandwidth for users watching the same content at the same time, as a single data stream can send packets to a large number of subscribers. For the broadcaster and the network, the marginal cost of an additional viewer is nil, whereas unicast distribution engenders additional costs with every new user.

Today, however, several obstacles still stand in the way of widespread IP multicasting. Traditional multicasting protocols have been seen as cumbersome and the configuration complex; there is no common multicast language, and only a small number of routers and devices are equipped to do it.

Given the “best effort” environment that persists widely over the Internet today, networks will need to be carefully designed to balance the quality of service issues of unreliable delivery and packet loss with the threat of packet duplication and resulting network congestion.

Video Pre-caching

In a nutshell

Switching to the device side, video pre-caching involves pre-staging video on-demand content on customer devices in preparation for future viewing. As hard drives on customer equipment such as set-top boxes are not full, it harnesses these extra resources to boost capacity during peak viewing times. Video content is sent to the home gateway at off-peak times to provision the subscriber’s network with preloaded content. Subscribers can then stream their favorite pre-cached show on any device within the home network.

Pre-positioning may involve caching only a portion of the content in the device – the beginning of a TV show, for example – to ensure quick startup and then switching seamlessly into unicast or multicast delivery when the viewer actually plays the video.

Suppliers and use cases

Examples of pre-staging include various intelligent content positioning systems that use predictive technologies to analyze and viewer behavior to predict future viewing choices (a task facilitated by binge watchers!). We can again cite Broadpeak’s nano-CDN technology, which may also be used to pre-place content within the home to deliver streams to household devices.

Pros and cons

Content pre-caching can greatly improve startup times and capitalize on lulls in traffic to reduce latency and take pressure off networks when users connect en masse.

Its potential to relieve networks, however, remains relatively limited, as user devices have only so much storage capacity, and as it is still difficult to predict exactly when and what a user will watch and have the updated content ready. Mobile meanwhile presents a unique set of challenges associated with networks constrained by spectrum.

All in all, optimizing storage capacity, type of content, frequency of positioning and updates, and compatibility with emerging network technologies are all major design focuses that must be optimized if we hope to build effective pre-caching systems.

LTE Broadcast

LTE broadcast diagrams

In a nutshell

Staying on the topic of mobile (and pre-caching), Long Term Evolution network technology is laying the foundation for innovative solutions to mobile video. While the technology offers a range of advantages to reduce latency between device and cell, improve connections and optimize battery life, we will concentrate solely on its video delivery capabilities.

LTE Broadcast provides an alternative to unicast distribution though broadcast (one-to-all) or multicast (one-to-many) distribution of content. With LTE, base stations can serve a very high number of users simultaneously, sending the same packet to every system within its network at the same time.

Suppliers and use cases

LTE offers possibilities for both VoD unicast-offloading and live stream delivery. In the first case, popular content can be pre-placed in devices’ caches via broadcast during network off-peak hours, as described above. For live delivery, premium live content and venue casting (instant replays, newscasts) are the two most often-cited use cases. Content is broadcasted in well-defined areas – stadiums, dense urban areas, regions or even entire countries – from either a single cell, or from cells forming single frequency network that can synchronize sending of the data over the same time slot. Ericsson has developed extensive offerings around LTE technologies.

Pros and cons

As a means to send information all at once, LTE provides a robust solution to last-mile issues in mobile networks. It can also be used in tandem with unicast and pre-caching systems, decreasing the amount of traffic required to send certain information.

Today, an increasing number of devices are equipped with LTE chipsets and the appropriate eMBSB-enabled (Evolved Multimedia Broadcast Multicast Service) middleware. However, this technology is still a ways from becoming standard. Customer devices still present storage and processing issues, while LTE requires installing additional antennas on base stations for data transmission, a significant startup cost.

Peer-to-peer

We obviously can’t talk about delivery solutions without touching briefly on peer-accelerated streaming. Like content pre-caching, peer-accelerated streaming harnesses the power of viewer devices to assist in video delivery. In this type of mesh network, users watching the same content at the same time can share data directly instead of systematically retrieving content from a CDN or server. It thus helps alleviate network congestion by decentralizing data exchanges. The more viewers there are to share the information, the better the system works, allowing peer-accelerated streaming to overcome the scaling limitations imposed by traditional server and CDN infrastructures. With an obvious bias, we believe that peer-accelerated streaming offers great potential for large live sporting events and popular VoD content.

Diagrams showing differencies with and without Peer to peer solution

Conclusion: towards hybrid delivery systems

It is clear that delivery architectures must innovate to keep pace with growing traffic demands and pressure on existing infrastructure. The examples above all demonstrate that one single technology will not be sufficient to rise to the challenges facing video delivery today. With content pre-caching and home gateway multicasting, we see an increasing trend towards involving client devices in the delivery architecture in a way that was hitherto largely unexplored.

In short, the solution of the future is a hybrid solution, one that combines various technologies that each have their place along the delivery chain and that are both interoperable and format agnostic. While we at Streamroot have focused on the possibilities offered by peer-to-peer, we firmly believe that ensuring quality, scalable delivery will only be possible through a collective effort.

Many thanks to the contributions of those companies cited in this post. Stay tuned for a more in-depth look at the delivery technologies featured here, how they can ensure more efficient video delivery, and their thoughts on the role of peer-accelerated streaming in this ecosystem!

delivery solutions: building the infrastructure of tomorrow


Scroll to Top