Skip to content
  • SVTA University Calendar
  • Courses
    • In-Person Training
  • Hot Topics
  • Education Resources
    • Conferences
      • Demuxed
      • Mile High Video
      • NAB Streaming Summit
      • SEGMENTS
      • Streaming Tech Sweden
    • Industry Resources
    • Media Samples
    • SVTA Webinars
  • Instructors
  • Register
  • Log In
  • SVTA University Calendar
  • Courses
    • In-Person Training
  • Hot Topics
  • Education Resources
    • Conferences
      • Demuxed
      • Mile High Video
      • NAB Streaming Summit
      • SEGMENTS
      • Streaming Tech Sweden
    • Industry Resources
    • Media Samples
    • SVTA Webinars
  • Instructors
  • Register
  • Log In
$0.00 0 Cart

Conference Proceedings

  • Home
  • Don’t let latency ruin your longtail: an introduction to “dref MP4” caching
Don’t let latency ruin your longtail: an introduction to “dref MP4” caching

Description

Many setups in large scale on-demand video streaming nowadays rely on a just-in-time packager to deliver remotely stored MP4 content. Unfortunately, the latency to such a storage backend can be suboptimal and a just-in-time packager needs to make a relatively high number of requests to the source content to package it dynamically. This will impact the overall performance of the packager, as well as the start-up delay that customers will experience when a stream that hasn’t been cached on the CDN is requested (i.e., longtail content).

In my talk, I will propose how the number of requests to the remote storage backend can be minimized by using a novel but spec compliant approach to packaging MP4. In essence, it’s about generating an additional MP4 that acts as an intermediary between the packager and the source content. This MP4 does not contain media data, but merely references the source content using MP4’s ‘dref’ box. What this ‘dref MP4’ does contain, is the index information (sample tables) of the original track, stored in a ‘moov’ box. This is the small but vital bit of information that a just-in-time packager needs to handle each and every incoming request. By omitting the original media data, a dref MP4 becomes very small and can be cached easily in a reverse proxy cache that sits between the just-in-time packager and the remote storage backend. Or it can be stored locally on the packager itself. Either way, this approach will ensure that the just-in-time packager will have much faster access to information that it needs to access very often. The impact of this approach is considerable. In my tests with Unified Origin, Unified Streaming’s just-in-time packager, an MPD (DASH client manifest) was generated 70x times faster (as it contains timelines for all track in a stream, a packager a will need to read the index information of each of those tracks). The gains when packaging a media segment were much smaller, but still quite significant with an improvement of 1.5x. Also, the general throughput of the packager improved by ~10-20%. In my talk I will explain the above by going into the nitty gritty details of the relatively high number of requests that Unified Origin make to a storage backend, explaining what each of these requests are for to give the audience a more thorough understanding of the inner workings of a just-in-time packager in general, and Unified Origin in specific, while also introducing them to the concept of a dref MP4, which – based on my experience thus far – will be a format that most have never heard of. With this unique blend of technical insight, practical gains and an introduction to a novel type of MP4, I think this talk will serve the audience at Demuxed particularly well. I hope you do, too. Presented at Demuxed 2020.

Conference

Demuxed 2020

Speakers

Boy van Dijk

Senior Solutions Engineer

Learning Categories

Encoding
Packaging
JIT Packaging
Storage

Other Proceedings

Here are some other proceedings that you might find interesting.

What Codec Should I Use?

Alan Resnick

Doing Server-Side Ad Insertion on Live Sports for 25.3M Concurrent Users

Ashutosh Agrawal

Is now the time to solve the deepfake threat?

Roderick Hodgson

Super Resolution: The scaler of tomorrow, here today!

Nick Chadwick

The do's and don'ts about Streaming security

Javier Brines Garcia

Modeling the conceptual structure of FFmpeg in JavaScript

Ryan Harvey

Objectionable Uses of Objective Quality Metrics

Richard Fliam

RTMP: web video innovation or Web 1.0 hack… how did we get to now?

Sarah Allen

Large-Scale Media Archive Migration to the Cloud

Konstantin Wilms

HEVC Upload Experiments

Chris Ellsworth

Related Courses

Below are some courses that might interest you based on the learning categories and topic tags of this conference proceeding.

What Codec Should I Use?

Alan Resnick

Doing Server-Side Ad Insertion on Live Sports for 25.3M Concurrent Users

Ashutosh Agrawal

Is now the time to solve the deepfake threat?

Roderick Hodgson

Super Resolution: The scaler of tomorrow, here today!

Nick Chadwick

The do's and don'ts about Streaming security

Javier Brines Garcia

Modeling the conceptual structure of FFmpeg in JavaScript

Ryan Harvey

Objectionable Uses of Objective Quality Metrics

Richard Fliam

RTMP: web video innovation or Web 1.0 hack… how did we get to now?

Sarah Allen

Large-Scale Media Archive Migration to the Cloud

Konstantin Wilms

HEVC Upload Experiments

Chris Ellsworth

Follow

Twitter Linkedin-in

User Area

  • Account
  • FAQs
  • Orders
  • Registration
  • Account
  • FAQs
  • Orders
  • Registration

Resources

  • About
  • FAQs
  • Legal Hub
  • Support
  • How-To Take A Course
  • How-To Navigate the Interface
  • About
  • FAQs
  • Legal Hub
  • Support
  • How-To Take A Course
  • How-To Navigate the Interface

SVTA Sites

  • Diversity and Inclusion
  • LABS
  • OATC
  • Open Caching
  • SEGMENTS
  • Streaming Video Wiki
  • SVTA Fellows
  • SVTA University
  • Diversity and Inclusion
  • LABS
  • OATC
  • Open Caching
  • SEGMENTS
  • Streaming Video Wiki
  • SVTA Fellows
  • SVTA University

© Copyright Streaming Video Technology Alliance (SVTA).

About the SVTA University

The SVTA University (SVTAU) is an educational arm of the Streaming Video Technology Alliance, providing courses and other instructional content related to understanding and working with components within the streaming video stack.

About the SVTA

The Streaming Video Technology Alliance is a global technical association committed to bringing video streaming companies together to help build a better viewer experience at scale. Find out more at www.svta.org.

Payment Forms

Stay In-the-Know!

Enter your email address below to subscribe to our newsletter for the latest in available courses and other Institute news. Note that by doing so, you agree to our privacy policy.

Loading...

Insert/edit link

Enter the destination URL

Or link to existing content

    No search term specified. Showing recent items. Search or use up and down arrow keys to select an item.