Media Transcoding With Backblaze B2 and Vultr Optimized Cloud Compute

Since announcing the Backblaze + Vultr partnership last year, we’ve seen our mutual customers build a wide variety of applications combining Vultr’s Infrastructure Cloud with Backblaze B2 Cloud Storage, taking advantage of zero-cost data transfer between Vultr and Backblaze. This week, Vultr announced Optimized Cloud Compute instances, virtual machines pairing dedicated best-in-class AMD CPUs with just the right amount of RAM and NVMe SSDs.

To mark the occasion, I built a demonstration that both showcases this new capability and gives you an example application to adapt to your own use cases.

Imagine you’re creating the next big video sharing site—CatTube—a spin-off of Catblaze, your feline-friendly backup service. You’re planning all sorts of amazing features, but the core of the user experience is very familiar:

  • A user uploads a video from their mobile or desktop device.
  • The user’s video is available for viewing on a wide variety of devices, from anywhere in the world.

Let’s take a high-level look at how this might work…

Transcoding Explained: How Video Sharing Sites Make Videos Shareable

The user will upload their video to a web application from their browser or a mobile app. The web application must store the uploaded user videos in a highly scalable, highly available service—enter Backblaze B2 Cloud Storage. Our customers store, in the aggregate, petabytes of media data including video, audio, and still images.

But, those videos may be too large for efficient sharing and streaming. Today’s mobile devices can record video with stunning quality at 4K resolution, typically 3840 × 2160 pixels. While 4K video looks great, the issue is that even with compression, it’s a lot of data—about 1MB per second. Not all of your viewers will have that kind of bandwidth available, particularly if they’re on the move.

So, CatTube, in common with other popular video sharing sites, will need to convert raw uploaded video to one or more standard, lower-resolution formats, a process known as transcoding.

Transcoding is a very different workload from running a web application’s backend. Where an application server requires high I/O capability, but relatively little CPU power, transcoding is extremely CPU-intensive. You decide that you’ll need two sets of machines for CatTube—application servers and workers. The worker machines can be optimized for the transcoding task, taking advantage of the fastest available CPUs.

For these tasks, you need appropriate cloud compute instances. I’ll walk you through how I implemented CatTube as a very simple video sharing site with Backblaze B2 and Vultr’s Infrastructure Cloud using Vultr’s Cloud Compute instances for the application servers and their new Optimized Cloud Compute instances for the transcoding workers.

Building a Video Sharing Site With Backblaze B2 + Vultr

The video sharing example comprises a web application, written in Python using the Django web framework, and a worker application, also written in Python, but using the Flask framework.

Here’s how the pieces fit together:

  1. The user uploads a video from their browser to the web app.
  2. The web app uploads the raw video to a private bucket on Backblaze B2.
  3. The web app sends a message to the worker instructing it to transcode the video.
  4. The worker downloads the raw video to local storage and transcodes it, also creating a thumbnail image.
  5. The worker uploads the transcoded video and thumbnail to Backblaze B2.
  6. The worker sends a message to the web app with the addresses of the input and output files in Backblaze B2.
  7. Viewers around the world can enjoy the video.

These steps are illustrated in the diagram below.

Click to enlarge.

There’s a more detailed description in the Backblaze B2 Video Sharing Example GitHub repository, as well as all of the code for the web application and the worker. Feel free to fork the repository and use the code as a starting point for your own projects.

Here’s a short video of the system in action:

Some Caveats:

Note that this is very much a sample implementation. The web app and the worker communicate via HTTP—this works just fine for a demo, but doesn’t account for the worker being too busy to receive the message. Nor does it scale to multiple workers. In a production implementation, these issues would be addressed by the components communicating via an asynchronous messaging system such as Kafka. Similarly, this sample transcodes to a single target format: 720p. A real video sharing site would transcode the raw video to a range of formats and resolutions.

Want to Try It for Yourself?

Vultr’s new Cloud Compute Optimized instances are a perfect match for CPU-intensive tasks such as media transcoding. Zero-cost ingress and egress between Backblaze B2 and Vultr’s Infrastructure Cloud allow you to build high performance, scalable applications to satisfy a global audience. Sign up for Backblaze B2 and Vultr’s Infrastructure Cloud today, and get to work!


About Pat Patterson

Pat Patterson is the chief technical evangelist at Backblaze. Over his three decades in the industry, Pat has built software and communities at Sun Microsystems, Salesforce, StreamSets, and Citrix. In his role at Backblaze, he creates and delivers content tailored to the needs of the hands-on technical professional, acts as the “voice of the developer” on the Product team, and actively participates in the wider technical community. Outside the office, Pat runs far, having completed ultramarathons up to the 50 mile distance. Catch up with Pat via Twitter or LinkedIn.