Cloud Hosting

This blog is co-authored by Francois Guilleautot, Director of cloud solutions, Ateme.

Since its introduction at the 1999 CES show in Las Vegas, Digital Video Recording (DVR) has existed in various forms. While the technology is not new (Ateme has offered DVR solutions since 2016), the transition to over-the-top (OTT) streaming has accelerated adoption, with the DVR market expected to grow from $6.4B USD in 2023 to $16.4B USD by 2030. Consumers now record more content than ever, and expect recordings to be available on any device, from anywhere. In fact, recording capability is now the most desired functionality for a streaming service.

Although demand for recording capabilities has increased with streaming, the concept itself is not new. Technology to record television for later viewing has been available for more than 50 years. In the age of analog TV, VHS and Betamax tape recorders were available to save copies of your favorite programs. Prior to that, television viewers may remember G-Code, VideoPlus+, or ShowView codes in the TV Guide to ease the pain of setting a recording time.

In the era of digital television, hard drives were integrated into set-top-boxes (STBs) to allow for recording capabilities. A large number of legacy STBs are still in use today. However, TV operators ultimately realized that STB DVR capabilities were inefficient and expensive. Video requires substantial storage capacity and high performance. Providing millions of users with hundreds of hours of storage means managing fleets of STBs with pricey, failure-prone, and rapidly aging hard drives.

To reduce management and services costs, Network Digital Video Recording (nDVR) platforms, such as Atemes NEA DVR, emerged. The concept is simple: rather than dispersed recordings over countless hard drives, providers merge their storage on a single unified platform. This improves efficiency and reduces overhead, leading to improved total cost of ownership for the operator. With nDVR, drives are part of an owned infrastructure with better protection and improved durability, while allowing customer access via the operators closed network or via the internet.

As operators improved operations, the viewers experience improved with recordings available from any device connected to the nDVR. Consumer began to record more content, in increasingly higher resolution, including 4K. One of Atemes customers scaled up to 300 racks of video storage for its nDVR system, accommodating content over 15 years of age.

Expanding video recording systems comes with a distinct set of difficulties. The first challenge is managing massive scalability. Modern streaming services generate anywhere from a few hundred gigabytes to tens of terabytes of recordings each day, depending on the legal status of the country hosting the platform. With Shared-copy permitted by most nations, if multiple users record the same content (e.g., a football match or a movie), a single copy is stored and accessed by all who record it. This shrinks the amount of raw storage space required by a factor of 10 or even 100. Even with shared copy techniques, network DVR platforms can expand to enormous scale, reaching hundreds of petabytes. Managing storage volumes of this magnitude requires substantial compute, network, and storage resourcesup to tens of thousands of HDDs or SSDs.

DVR places an incredible amount of pressure on storage. To ensure an optimal viewing experience (start time, resolution, re-buffering), high Inputs/Outputs Operations per Second (IOPS) are required to support throughput traffic on the storage. Commodity, high-capacity hard drives do not have the IOPS required to deliver content to thousands or tens of thousands concurrent viewers requesting high-resolution content simultaneously. Using more drives in parallel improves storage performance, but requires additional CPU capacity to fully use extra storage, which increases cost.

Beyond the sheer size and performance requirements for storage, availability (and therefore redundancy) is an important consideration. Subscribers expect their content to be available instantly, even months or years after the recording. To allow such a high level of availability, nDVR vendors use high-performance Network-Attached Storage (NAS) with large redundant arrays of independent disks (RAID), or a distributed storage system with a dedicated erasure coding mechanism for better performance and density. However, these systems have limited scalability. The additional storage requires matching compute power and networking to ensure smooth operation, compounding already difficult storage capacity requirements.

Such a large, high-performance video storage platform comes with high maintenance costs. Large recording platforms require hundreds to thousands of MWh per year in power and cooling, as well as engineers dedicated to the hardware management who may encounter frequent disk failures and other hardware issues across hundreds of servers. Engineers must regularly roll out operating system and software updates to fix bugs and avoid ransomware attacks. A recent example is the 2021 LOG4J Zero-Day vulnerability.

Migrating DVR platforms to the cloud provides an elegant solution to the problems previously listed, namely scalability, operational complexity, cost, and security. For example, Amazon Simple Storage Service (S3) offers virtually unlimited storage capacity with high-speed, built-in data transfers, and data redundancy by writing across multiple Availability Zones within a region. This provides 99.99% availability and 11-9s of durability for recordings. Adding a geo-redundancy dimension, not economically viable with on-premises platforms, improves recording availability. On top of offering better reliability, Amazon Web Services (AWS) also handles the undifferentiated heavy lifting of hardware and OS security. AWS offers additional security features such as encryption at rest and in transit, and integration with AWS Identity and Access Management (IAM) for fine-grained access control of video content.

The major benefit of migrating to Amazon S3 cloud storage is native storage tiering. Storage tiering is a feature that Ateme offers with its legacy nDVR system and that AWS offers natively. This allows you to choose from different storage classes based on data access patterns, with less frequently accessed video files located to lower-cost storage tiers, optimizing costs without compromising accessibility. This is especially relevant for video recordings as usage patterns change over time, with older recordings requested less often than recent recordings.

Ateme uses four Amazon storage services to optimize cost over the course of an asset lifecycle. In an end-to-end cloud-native OTT platform, an asset starts its lifecycle on premises or on Amazon gp2 Elastic Block Storage (Amazon EBS) as part of a live channel rolling buffer for time-shifted TV (TSTV) for maximum performance (IOPS). The asset then moves to an S3 Standard bucket to enable Catch-up or backwards EPG services. As the asset ages out of the backward EPG, its popularity and number of requests decreases. With a decreasing number of requests, it will be moved to S3 Infrequent Access and then to Glacier Instant Retrieval (GIR). GIR is the lowest tier of storage available to reduce prices while still allowing for instant playback. S3 Intelligent tiering allows assets to be programmatically moved between storage classes based on access patterns.

To best leverage the benefits of cloud storage, Ateme rearchitected its entire recording pipeline. In 2023, it launched a new cloud-native recording platform, NEA Genesis. Ateme designed NEA Genesis using a micro-services architecture to operate natively on AWS. This move away from a monolithic architecture allows Ateme to scale out logical blocks independently such as ingest, storage, and playout.

Ateme NEA Genesis High Level Architecture

Scaling logical blocks independently allows streaming platforms to optimize their non-linear video operations with a flexible and unified solution for VOD and recordings. With NEA Genesis, not only is it possible to grow storage independently of compute, but it is now possible to scale up ingest and playout dynamically. This provides flexibility to add large batches of VOD and event-based channels, or scale up playout capacity when demand surges. On-premises infrastructure scales these resources together, and typically only allows scaling up, which leads to over-provisioning. AWS infrastructure supports Atemes ability to scale ingest, storage, and egress independently, a major advantage over the limitation of on-premises hardware.

NEA Genesis includes additional modules for advanced functionality. For example, it supports encrypted or clear asset storage, with a dedicated packager/re-packager and encryption service to ensure platform compatibility with future standards. This means that assets can be ingested in a single format (e.g., DASH or CMAF) and redistributed in whatever combination of streaming protocol and DRM that is relevant in the future. NEA Genesis customers also benefit from both the advanced features of Atemes in-house built packaging engine and Atemes extensive integration library with CMS and DRM partners for frictionless onboarding.

To learn more about the benefits of cloud DVR solutions from Ateme, visit the companys website.

See the original post:
Cloud DVR levels up with video expertise and operational agility from Ateme and AWS | Amazon Web Services - AWS Blog