Cloud DVR: Opportunities and Challenges
A look back a few decades at the video recording industry proves that history has the tendency to repeat itself. Analogue magnetic tape formats such as Sony's Betamax and JVC's VHS slowed in the late 1990s under competition from personal video recorders (PVRs) incorporating digital disk drives. As of Q1 2015, PVRs have penetrated 49% of U.S. TV households. Global PVR sales have reached a plateau with low single-digit growth expected in the coming years. One reason for the decline in sales is the increasing popularity of viewing television content on portable devices.
Cloud DVR is the logical successor to the PVR. The technology is ready, and a growing number of television viewers are finding this new service very appealing. Cloud DVR frees viewers from the limited storage capacity of PVRs by allowing them to record a potentially unlimited number of programs, simultaneously if they wish, and replay them on any device of their choice at any time.
Cloud DVR is a great opportunity for Pay TV, cable, and telecom operators in their quest for convergence to attract new subscribers, reduce churn, and increase the average revenue per user.
Classic Cloud DVR Architecture
Each Cloud DVR video stream starts its journey encoded by the provider of the TV channel. Commercials are typically inserted by a splicing device on a national or geographic basis. Once received at the operator's premises, transcoding is applied to transform the incoming content to the proper format and bitrates for ABR distribution before the stream is delivered to the operator's network.
At this stage, packaging can prepare the content for all possible formats and resolutions required for all possible viewing devices. Preparation is now complete. All combinations of codec, DRM, format, and bitrates are then recorded in a large and high-performance central data store. When requested by the subscriber device, the ABR or IPTV contents are delivered by a streaming server to the consumer devices. This delivery is performed through the operator network or via the internet and cached in the content delivery network (CDN) in order to off-load central systems and increase performance.
The Cloud DVR service is almost in place. The service platform will be the interface between this infrastructure and each user. It manages the recording’s information, the user’s quota, interaction with the electronic program guide, and the digital rights management licence delivery. Security is key. Encryption can be applied in different places depending on the system architecture. In the more conservative case, encryption is performed during initial recording. Encrypted content is then decrypted on the user device. It could also be applied at playout, but this requires other secure-storage mechanisms.
Optimized Cloud DVR Infrastructure
Both the technology underlying Cloud DVR and the Cloud DVR component characteristics themselves have evolved a lot in the past year or two. Let's look more closely and try to understand how they influence the Cloud DVR architecture and therefore how this creates an opportunity to rework and optimize the architecture.
Private Copy vs. Shared Copy
In the private-copy model, every time a customer records a program, the Cloud DVR system creates and stores a new file that contains this piece of content. This asset is exclusive to this user and cannot be shared with other users. Each recording will be duplicated and stored as many times as requested.
In the shared-copy model, when the start and end time of a recording are similar (typical of event-based recordings), a single copy of this content is stored and shared. Recordings are not duplicated, resulting in a massive optimization of storage space. To the extent that this model is adopted, and therefore possible in the eyes of the law, this system naturally seems to prevail. But that is only in theory. In practice, some deviations come into effect. Any difference between start and end times, caused by errors in the EPG or instant-based recording behavior for example, and this shared-copy model becomes technically a private-copy model.
A middle path exists: the true-shared model. More in tune with reality, this model entails pooling segments of videos that are common to all recordings, whatever their start or end time. It guarantees operators efficient optimization of their storage.
Due to an uncertain legal environment and a rather defensive approach from content owners with respect to this innovation, the deployment of Cloud DVR services requires a platform able to fine-tune the model at the channel or program level, depending on the regulatory situation. Some programs or channels may be eligible for shared-copy, while others are only eligible for private copy. The core system should manage this complexity and has to be flexible enough to let this strategy evolve over time.
The promise of a lifetime recording library demands a future-proof approach capable of operating in a quickly changing technology landscape.
Considering the large number of new viewing devices flooding the markets, the constant flip-flopping of formats, and the introduction of new DRMs, this unpredictability has put some pressure on Cloud DVR platforms. Assets recorded today need to be readable in the future on all sorts of devices, including those that have not yet been developed.
One approach entails reprocessing the entire catalog of assets when a new format arises, and extending de facto the storage space to accommodate an additional version of each recording. Another option favours a single version of each content stored in a mezzanine format (or pivot format) which is repackaged to the requested format at delivery time. This repackaging has to be performed after storage. More conveniently, it could be combined with playout in an origin server providing real-time repackaging, encryption, and playout.
The savings are massive: Moving the repackaging downstream of the storage will reduce the storage burden by at least 2 to 3 times depending on the number of formats and DRM combinations. It will also reduce the storage throughput requirements by an equal amount.
Content caching allows one step further towards cost optimisation. This involves caching the most popular content in a CDN. A relative minority of popular shows typically count for 80% of total viewing figures, so this approach reduces the throughput requirements of the central storage.
To optimise delivery costs further, packaging may be performed closer to the user: the mezzanine format will be delivered once, cached into the network as close as possible to the user, and packaged and delivered on request.
When the same content in the same format and same DRM is required by a user connected to the same node, it will be served by the edge cache. Content not already cached will be served by the previous node and will be cached at the first time it is served. Deploying caches at the edge will offload the long distance network. All formats and DRM variants of the content will still need to be carried along the network, even if only to feed the edge cache once and for all. Moving the repackaging to the edge cache will reduce the network load even more as only the mezzanine version of the content will be moved to the edge. In this latter case, the encryption will have to be performed at the edge level for the same reason as previously explained.
Some important elements have to be considered when designing an optimised system:
- Packaging on the edge implies encrypting at the edge. This could be a limitation unless the content is secured along the entire delivery chain with encryption performed all along the value chain, but it provides an opportunity for session-based encryption.
- The deployment and deliverability of all nonlinear services including Cloud DVR with chunk caching.
- Arbitrate between storage and caching costs where storage is becoming more cost effective over time.
[Damien Lucas is chief technical officer of ANEVIA. Streaming Media accepts vendor-contributed articles such as this one based solely on their value to our readers.]
Though fraught with challenges, the move to the cloud for DVR, UI-UX, and ad insertion is helping service providers and operators maximize revenue and deliver a better experience to their subscribers.