Wednesday, June 24, 2015

Building a mobile video delivery network? part III


Content providers and aggregators have obviously an interest (and in some case a legal obligation) to control the quality of the content they sell to a consumer. Without owning networks outright to deliver the content, they rent capacity, under specific service level agreements to deliver this content with managed Quality of Experience. When the content is delivered over the “free” internet or a mobile network, there is no QoE guarantee. As a result, content providers and aggregators tend to “push the envelope” and grab as much network resource as available to deliver a video stream, in an effort to equate speed and capacity to consumer QoE. This might work on fixed networks, but in mobile, where capacity is limited and variable, it causes congestion.

Obviously, delegating the selection of the quality of the content to a device should be a smart move. Since the content is played on the device, this is where there is the clearest understanding of instantaneous network capacity or congestion. Unfortunately, certain handset vendors, particularly those coming from the consumer electronics world do not have enough experience in wireless IP for efficient video delivery. Some devices for instance will go and grab the highest capacity available on the network, irrespective of the encoding of the video requested. So, for instance if the capacity at connection is 2Mbps and the video is encoded at 1Mbps, it will be downloaded at twice its rate. That is not a problem when the network is available, but as congestion creeps in, this behaviour snowballs and compounds congestion in embattled networks.
As more and more device manufacturers coming from the computing world (as opposed to mobile) enter the market with smartphones and tablets, we see wide variations in the implementation of their native video player.
Consequently, operators are looking at way to control video traffic as a means to maybe be able to monetize it differently in the future. Control can take many different aspects and rely on many technologies ranging from relatively passive to increasingly obtrusive and aggressive.

In any case, the rationale for implementing video control technologies in mobile networks goes beyond the research for the best delivery model. At this point in time, the actors have equal footing and equal interest in preserving users QoE. They have elected to try and take control of the value chain independently. This has resulted in a variety of low level battles, where each side is trying to assert control over the others.
The proofs of these battles are multiple:
  • Google tries to impose VP9 as an alternative to H.265 /HEVC: While the internet giant rationale to provide a royalty-free codec as the next high efficiency codec seems innocuous to some, it is a means to control the value chain. If content providers start to use VP9 instead of H.265, Google will have the means to durably influence the roadmap to deliver video content over the internet.
  • Orange extracts peering fees from Google / YouTube in Africa: Orange as a dominant position for mobile networks and backhaul in Africa and has been able to force Google to the negotiating table and get them to pay peering fee for delivering YouTube over wireless networks. A world’s first.
  • Network operators implement video optimization technologies: In order to keep control of the OTT videos delivered on their networks, network operators have deployed video optimization engine to reduce the volume of traffic, to alleviate congestion or more generally to keep a firmer grip on the type of traffic transiting their networks.
  • Encryption as an obfuscation mechanism: Content or protocol encryption has traditionally been a means to protect sensitive content from interception, reproduction or manipulation. There is a certain cost and latency involved in the encoding and decoding of the content, so it has remained mostly used for premium video. Lately, content providers have been experimenting with the delivery of encrypted video as a means to obfuscate the traffic and stop network operators from interfering with it.
  • Net neutrality debate, when pushed by large content providers and aggregators is oftentimes a proxy for commercial battle. Th economics of the internet have evolved from browsing to streaming and video has disrupted the models significantly. The service level agreements put in place by the distribution chains (CDNs, peering points...) are somewhat inadequate for video delivery.


We could go on and on listing all the ways that content providers and network operators are probing each other’s capacity to remain in control of the user’s video experience. Ultimately, these initiatives are isolated but are signs of large market forces trying to establish dominance over each other. So far, these manoeuvres have reduced the user experience. The market will settle in a more collaborative mode undoubtedly as the current behaviour could lead to mutually assured destruction. The reality is simple. There is a huge appetite for online video. An increasing part of it takes place on mobile devices, on cellular networks. There is money to be made if there is collaboration, the size of the players is too large to establish a durable dominance without vertical integration.

Tuesday, June 23, 2015

Building a mobile video delivery network? part II


Frequently, in my interactions with vendors and content providers alike, the same questions are brought up. Why aren’t content providers better placed to manage the delivery of the content they own rather than network operators? Why are operators implementing transcoding technologies in their networks, when content providers and CDN have similar capabilities and a better understanding of the content they deliver? Why should operators be involved in controlling the quality of a content or service that is not on their network?

In every case, the answer is the same. It is about control. If you look at the value chain of delivering content over wireless networks, it is clear that technology abounds when it comes to controlling the content, its quality, its delivery and its associated services at the device, in the network, in the CDN and at the content provider. Why are all the actors in the delivery chain seemingly hell-bent on overstepping each other’s boundary and wrestle each other’s capacity to influence content delivery?

To answer this question, you need to understand how content used to be sold in mobile networks. Until fairly recently, the only use case of “successful” content being sold on mobile networks was ringtones. In order to personalize your phone, one use to go to their operator’s portal and buy a ringtone to download to one’s device. The ringtones were sold by the operator, charged on one’s wireless bill, provided by an aggregator, usually white-labelled who would receive a percentage of the sale, and then kick back another percentage of their share to the content provider itself who created the ringtone.
That model was cherished by network operators. They had full control of the experience, selecting themselves the content aggregator, in some case the content providers, negotiating the rates from a position of power, and selling to the customer under their brand, in their branded environment, on their bills.

This is a long way from today’s OTT, where content and services are often free for the user, monetized through advertisement or other transparent scheme, with content selected by the user, purchased or sourced directly on the content provider’s site, with no other involvement from the network operator than the delivery itself. These OTT (Over-The-Top) services threaten the network operator’s business model. Voice and messaging are the traditional revenue makers fro operators and are decreasing year over year in revenue, while increasing on volume due to the fierce competition of OTT providers. These services remain hugely profitable for networks and technology has allowed great scalability with small costs increments, promising healthy margins for a long while. Roaming prices are still in many cases extortionate. While some legislators are trying to get users fairer prices, it will be a long time before they disappear altogether.

Data, in comparison, is still uncharted territory. Until recently, the service was not really monetized, used as an appeal product to entice consumers to sign for longer term contracts. This is why so many operators initially launched unlimited data services. 3G, and more recently LTE have seen the latest examples of operators subsidizing data services for customer acquisition.

The growth of video in mobile networks is upsetting this balance though. The unpredictability and natural propensity of video to expand and monopolize network resources makes it a more visible and urgent threat as an OTT service. Data networks have greatly evolved with LTE with better capacity, speed and latency than 3G.  But the price paid to increase network capacity is still in the order of billions of dollars, when one has to take into account spectrum, licenses, real estate and deployment. Unfortunately, the growth in video in term of users, usage and quality outstrips the progress made in transport technology. As a result, when network operators look at video compounded annual growth rate exceeding 70%, they realize that serving the demand will continue to be a costly proposition if they are not able to control or monetize it. This is the crux of the issue. Video, as part of data is not today charged in a very sophisticated manner. It is either sold as unlimited, as a bucket of usage and/or speed. The price of data delivery today will not cover the cost of upgrading network capacity in the future if network operators cannot control better video traffic.

Additionally, both content providers and device vendors have diametrically opposed attitude in this equation. Device manufacturers, mobile network operators and content providers all want to deliver the best user experience for the consumer. The lack of cooperation between the protagonists in the value chain results paradoxically in an overall reduced user experience.


Wednesday, June 10, 2015

Google's MVNO - Project Fi is disappointing

A first look at Google's MVNO to launch in the US on the Sprint and T-Mobile networks reveals itself a little disappointing (or a relief if you are a network operator). I had chronicled the announcement of the launch from Mobile World Congress and expected much more disruption in services and pricing than what is announced here.

The MVNO, dubbed project Fi, is supposed to launch shortly and you have to request an invitation to get access to it (so mysterious and exciting...).

At first glance, there is little innovation in the service. The Google virtual network will span two LTE networks from different providers (but so is Virgin's in France for instance) and will also connect "seamlessly" to the "best" wifi hotspot. It will be interesting to read the first feedback on how the device selects effectively the best signal from these three options and how dynamically that selection occurs. Handover mid call or mid data sessions are going to be an interesting use case, Google assures you that the transition will be "seamless".

On the plus side, Google has really taken a page from Iliad's free disruptive service launched in France and one-time rumored to acquire T-Mobile US. See here the impact their pricing strategy  has had on the French telecommunications market.
  1. Fi Basic service comes with unlimited US talk and text, unlimited international text and wifi tethering for $20 per month.
  2. The subscriber is supposed to set a monthly data budget, whereas he selects a monthly amount and prepays 10$ per GB. At the end of the month, the amount of unused data is credited back for 1c / MB towards the following month. The user can change their budget on a monthly basis. Only cellular data is counted towards usage, not wifi. That's simple, easy to understand and after a little experimentation, will feel very natural.
  3. No contract, no commitment (except that you have to buy a 600+$ Nexus phone).
  4. You can send and receive all cellular texts and calls using Google hangouts on any device.
  5. Data roaming is same price as domestic but... see drawbacks

Here are, in my mind, the biggest drawbacks with the service as it is described.
  1. The first big disappointment is that the service will run initially only on Google's Nexus 6. I have spoken at length on the dangers and opportunities of a fully vertical delivery chain in wireless networks and Google at first seems to pile up the drawbacks (lack of device choice) with little supposed benefits (where is the streamlined user experience?).
  2. "Project Fi connects you automatically to free, open wifi networks that do not require any action to get connected". I don't know you, but I don't think I have ever come across one of these mysterious hotspots in the US. Even Starbucks or MC Donald free hot spots require to accept terms and conditions and the speed is usually lower than LTE. 
  3. Roaming data speed limited to 256 kbps! really? come on, we are in 2015. Even if you are not on LTE, you can get multi Mbps on 3G / HSPA. Capping at that speed means that you will not be streaming video, tethering or using data hungry apps (Facebook, Netflix, Periscope, Vine, Instagram...). What's the point, at this stage, better say roaming only on wifi  (!?).
In conclusion, it is an interesting "project", that will be sure to make some noise and have an impact on the already active price war between operators in the US, but on the face of it, there is too little innovation and too much hassle to become a mass market proposition. Operators still have time to figure out new monetization strategies for their services, but more than ever, they must choose between becoming wholesaler or value added providers.

Monday, June 8, 2015

Data traffic optimization feature set

Data traffic optimization in wireless networks has reached a mature stage as a technology . The innovations that have marked the years 2008 – 2012 are now slowing down and most core vendors exhibit a fairly homogeneous feature set. 

The difference comes in the implementation of these features and can yield vastly different results, depending on whether vendors are using open source or purpose-built caching or transcoding engines and whether congestion detection is based on observed or deduced parameters.

Vendors tend nowadays to differentiate on QoE measurement / management, monetization strategies including content injection, recommendation and advertising.

Here is a list of commonly implemented optimization techniques in wireless networks.
  •  TCP optimization
    • Buffer bloat management
    • Round trip time management
  • Web optimization
    • GZIP
    •  JPEG / PNG… transcoding
    • Server-side JavaScript
    • White space / comments… removal
  • Lossless optimization
    • Throttling / pacing
    • Caching
    • Adaptive bit rate manipulation
    • Manifest mediation
    • Rate capping
  • Lossy optimization
    • Frame rate reduction
    • Transcoding
      • Online
      • Offline
      • Transrating
    • Contextual optimization
      • Dynamic bit rate adaptation
      • Device targeted optimization
      • Content targeted optimization
      • Rule base optimization
      • Policy driven optimization
      • Surgical optimization / Congestion avoidance
  • Congestion detection
    • TCP parameters based
    • RAN explicit indication
    • Probe based
    • Heuristics combination based
  • Encrypted traffic management
    • Encrypted traffic analytics
    • Throttling / pacing
    • Transparent proxy
    • Explicit proxy
  • QoE measurement
    • Web
      • page size
      • page load time (total)
      • page load time (first rendering)
    • Video
      • Temporal measurements
        • Time to start
        • Duration loading
        • Duration and number of buffering interruptions
        • Changes in adaptive bit rates
        • Quantization
        • Delivery MOS
      • Spatial measurements
        • Packet loss
        • Blockiness
        • Blurriness
        • PSNR / SSIM
        • Presentation MOS


An explanation of each technology and its feature set can be obtained as part of the mobile video monetization report series or individually as a feature report or in a workshop.