How Synamedia delivers an optimized on-demand viewing experience with Cloud DVR on AWS

This post was co-authored by Meir Lehrer, Senior Director Video Network Product Management, Synamedia.

As the TV viewership experience has become more personalized and ubiquitous, service providers must make sure their subscribers’ favorite shows are available to watch anytime and anywhere. A modern Cloud-based Digital Video Recording (cDVR) solution becomes a critical feature for Multichannel Video Programming Distributors (MVPDs) to maximize the value of their licensed content. Scaling these cDVR services to meet the ever-growing demand on traditional data centers requires systematic investment for capacity expansion, as well as constant maintenance to ensure customer data is always preserved.

Enter Synamedia’s new Common Copy cDVR v2.0 technology. It was developed using Amazon Simple Storage Service (Amazon S3) as a cost effective cDVR solution that scales on-demand and eliminates the undifferentiated heavy lifting of managing large-scale storage arrays. Synamedia’s cDVR technology can reduce costs by over 95% compared with traditional DVR solutions.

A leading video technology provider for over 30 years and AWS ISV Partner, Synamedia is trusted by over 200 service providers worldwide. They have been at the forefront of delivering innovative cDVR solutions to some of the world’s largest video service providers.

This blog post explores how Synamedia addresses the challenges of scaling Cloud DVR while optimizing its cost.

Challenges with traditional Cloud DVR

Initially, Digital Video Recording (DVR) services were implemented by adding storage capacity to the subscriber’s set top box. This increases the cost of acquiring new subscribers and has a higher maintenance cost. Spinning disks running in uncontrolled environments and high rates of input and output operations have a low mean-time-to-failure, eventually requiring on-site technician visits to replace the device. In addition, recordings can only be consumed locally, storage capacity is limited, and data resiliency is fragile, since there is no backup strategy.

The second wave of DVR solutions were provided by in-network resources, which became known as Cloud DVR—the cloud, referring to the capabilities provided by service providers and deployed on-premises. This approach solves some issues previously described but comes with the burden of managing large storage arrays. MVPDs providing Cloud DVR were forced to invest in petabyte-scale storage arrays, estimating their capacity upfront with at least a few years of anticipation. They end up getting caught in cycles of hardware refresh and capacity expansion. Each of these cycles introduced newer generation hardware, increasing maintenance complexity. There are also power and cooling considerations, as those will be incurred for the fully deployed capacity, whether the DVR solution is being used or not. The on-premises system also demands constant maintenance to keep operating properly. Activities such as operational system and software updates, disk replacement, and enforcing a strong security posture take hours of expert IT human resources that could instead be focused on driving business goals. All these factors add up to a higher total cost of ownership of the Cloud DVR initiative.

In contrast, running Cloud DVR (cDVR) on AWS offers a more flexible and convenient approach to recording and accessing television content. Instead of relying on physical hardware or on-premises resources, cDVR on AWS leverages cloud-based storage and computing resources to store and deliver recorded content. This means that users access their recorded programs from virtually any device with an internet connection, whether it’s a smartphone, tablet, computer, or smart TV.

cDVR on AWS streamlines operations by removing the undifferentiated heavy lifting of maintaining storage arrays. AWS does the heavy lifting of data center operations like racking, stacking and powering servers. Service providers can free themselves from complex storage refresh and expansion projects by scaling on-demand with virtually unlimited storage capacity, and ultimately innovate with over 200 AWS services and features.

Cost-effective Cloud DVR on AWS with Synamedia

To optimize cDVR costs, Synamedia’s Common Copy cDVR v2.0 technology leverages multiple Amazon S3 storage classes. Storage classes such as Standard Infrequent Access (IA) and Glacier Instant Retrieval provide cost effective storage for infrequently or occasionally accessed recordings. Synamedia found that 20% or fewer of video recordings are viewed after three days, and the percentage reduces even further as content ages, reaching a small percentage after 14 days. Using multiple S3 storage classes allows Synamedia to automatically select the appropriate storage tier based on the age of the recording and reduces costs when compared to using the Standard storage class alone. Using this method also provides economic advantages to operators currently deploying on-premises storage systems, since a significant percentage of storage can be served by a cloud-based system instead.

Amazon S3 charges a small fee for requests against objects. Because of the nature of video streaming, the number of objects can grow immensely. For example, a single 1-hour television episode could have 1,800 2-second segments (objects) that need to be stored for just one profile. To reduce the number of requests required (such as HTTP data writes or retrievals, PUTs and GETs to and from S3) and improve network performance, Synamedia implements a concatenation buffer algorithm that groups batches of requests together at the source. By compressing and batching video segments at the source, the required number of requests is reduced by 1000s by the buffer algorithm in the common-copy recorder. As an example, 10 profiles of 2-second segments before optimization would require 18,000 PUT requests per hour (10 profiles / 2 second segments * 3,600 seconds). After optimization, with a concatenation window of 60 seconds, the number of requests is reduced to 60 PUTs per hour (1 / 60 * 3,600). When concatenating with a 15-minute window, this is reduced further to 4 PUTs per hour (1 / (15*60) * 3,600).

This concatenation buffer implementation is shown in the diagram below:

Fig 1: Synamedia common copy cDVR recorder architecture

Making such operational improvements requires an end-to-end view of the cDVR workflow. Synamedia designed a solution to ensure overall consumer experience would never be degraded.

The AWS architecture for Synamedia’s cDVR solution is shown in the following diagram:

Fig 2: Synamedia Virtual Digital Content Manager high-level architecture

Typically, cDVR is employed to record live linear television on a schedule. These live linear channels can either be deployed in AWS, or a hybrid Virtual Digital Content Manager (vDCM) model can be used, where the channels are deployed on-premise. In both the AWS and hybrid models, the recordings are durably persisted using Amazon S3. Amazon S3 provides durable object storage and is designed to exceed 99.999999999% (11 nines) data durability.

The key components of the architecture are:

1. Live linear channel processing is orchestrated using Amazon Elastic Kubernetes Service (EKS). Amazon EKS simplifies the orchestration of the containerized vDCM components, providing highly available and fault tolerant compute.
   1. vDCM Encoder encodes/transcodes ingested video to multiple video profiles with different resolutions and bitrates. The profiles compose the Adaptative Bitrate ladder that will be available for the video client.
   2. vDCM Linear Packager segments Adaptive bitrate (ABR) video profiles and provides push packaging to origin.
   3. vDCM Media Scalable Origin caches video segments to be pulled by just-in-time packager and common copy cDVR recorder.
   4. vDCM Just-in-time packager is used for live, video-on-demand and cDVR.
2. The Recorder Router (RR) component provides control plane communication to manage recording commands. The common copy cDVR recorder (CCR) stores the recordings in Amazon S3. The CCR is responsible for efficiently performing HTTP requests to ingest and retrieve data from storage:
   1. The recorder pulls segments from the media scalable origin (MSO).
   2. Based on a configuration setting, the recorder builds a buffer of all segments across all profile ladders for the required time-window.
   3. The recorder will store the segments until it reaches the time-window (e.g. 15 minutes). The recorder then performs a single PUT request of the buffered data.
   4. To resolve any latency effects, the just-in-time packager pulls from the MSO for the recording to be immediately available (rather than waiting to pull from the recorder’s storage).
3. Amazon S3 buckets leverage a combination of Intelligent Tiering and Lifecycle configuration to route video content automatically to the most cost-effective storage class based on changing access patterns. Cross-Region Replication can be optionally used to provide data redundancy for disaster recovery.
4. The Video Service Manager is an integrated monitoring and control management system that provides a single interface dashboard control of all Synamedia components across the network.

The just-in-time packager within the vDCM solution makes direct S3 GET requests to retrieve data via byte-range fetches. This is done as the concatenation buffer used to PUT content generates a single, combined large file comprising multiple segments The byte-range fetch enables a more efficient read mechanism by fetching only the segment data required.

Business impact

Synamedia’s concatenation buffer and built-in advanced management of Amazon S3 storage tiers resulted in the reduction from over $8,000 per month in requests, lifecycle transition, monitoring, and automation costs to less than $100 for one customer deployment. A 98.75% reduction in costs. This optimization requires a spatial and temporal view of the video streaming through the system. Modern ABR encoded video typically contains a ladder of 6 to 12 video, audio, and data profiles. Handling all these video segments across seconds and minutes to create the concatenation buffer, while still maintaining a system design adhering to three-second or less low-latency streaming, requires a careful design, implementation, and testing process.

Customers operating the Synamedia cDVR solution on AWS can lower their total cost of ownership while avoiding an increase in customer cost.

Conclusion

Synamedia’s cloud DVR solution on AWS helps video service providers deliver a superior on-demand viewing experience to customers while optimizing operational costs at scale. By appropriately managing capacity across different Amazon S3 storage classes, batching requests, and optimizing workflows, Synamedia minimizes cloud infrastructure expenses without compromising performance or user experience.  Synamedia’s three decades of video technology leadership, combined with AWS’s scalable and cost-effective cloud services, enable a best-in-class cloud DVR offering.

Start your journey towards an optimized and highly scalable cloud DVR solution today. Contact Synamedia to discuss how their cloud DVR on AWS can transform your video delivery capabilities while driving operational efficiency.