Cloudinary Blog

Building and scaling a service for developers - The bootstrapped way

Building and scaling a cloud service for developers

Last month I was invited to speak at Daho.am, Munich's developers conference. This conference was organized by Stylight, a very successful fashion technology startup.  Stylight signed up for a free Cloudinary account about 3 years ago and similarly to Cloudinary back then, Stylight were quite a young startup. Since then both companies have grown impressively together and Stylight are now a premium customer of Cloudinary, managing hundreds of millions of images.

At this conference I talked about building and scaling a B2D service (Business to Developers). Specifically, how to build and scale such a service for developers in the bootstrapped way with zero external funding, which is the Cloudinary story.

Below you can find the slides of the presentation and a full video of the talk from the conference. The presentation shares the insights we have from our experience of building a cloud service for developers and from growing the service and company while running, and in the organic way. In the presentation you can also find some behind-the-scenes details about the evolution of our internal architecture and some interesting numbers.

Full video of the presentation:

Presentation slides:

And here are some highlights of the presentation and the story.

Cloud service for developers

Before founding Cloudinary, we worked on building dozens of web applications for both startups and bigger companies for years. We felt that the R&D work involved in managing images was tremendous, especially compared to its proportional role in the product requirements of web and mobile applications. We wanted to help developers and eliminate all R&D work involved in managing images, from uploading, through resizing, cropping and manipulation to fast worldwide delivery. And if developers would be willing to pay for this service, that's nice too :-)

We quietly launched Cloudinary's image management service about 3.5 years ago. We've built a cloud-based API for uploading images and a URL-based API for on-the-fly image manipulation and delivery.

Quite quickly we realized that having a cloud API, as good as it can be, is not enough. In order for developers to use it, you need to have SDKs (client libraries) for all popular development frameworks such as Ruby on Rails, PHP, Java, .Net, Node.js, jQuery, iOS, Android and others. Developers look for one-liners of code to include in their applications in order to perform image upload, or for the embedding of manipulated and optimized images in their HTML pages. The tighter the framework integration is, the fewer the code changes that are required, and the bigger the chances of a successful integration.

Framework integration SDKs need to be accompanied by comprehensive documentation. Framework specific documentation is also very important as it allows developers to read instructions in the terminology they are used to, and find code samples in their programming language of choice instead of guessing the correct syntax based only on reference documentation or the sample code of other frameworks.

Creating documentation pages, tutorials and technical blog posts also had a great side effect: a lot of organic traffic comes from the Google searches of developers. This way the developers find a solution for their search queries, learn about our service, sign-up to the free service and eventually become happy paying customers. This marketing oriented side-effect was a great fit to Cloudinary's bootstrap approach that didn't involve a significant marketing budget in the early days of the company.

Instead of converging into a few leading frameworks, the development work seems to spread into more and more frameworks. If you have a cloud-based API and you count on the SDKs to ensure a successful integration, you need to monitor the trends of the frameworks. For example, see the chart below. It shows the percentage of Cloudinary's customers that used Ruby on Rails and Node.js in the last 2 years. As you can see Node.js popularity among our customers is growing rapidly and surpassed Rails a few months ago. By the way, AngularJS is the fastest growing framework among Cloudinary's customers.

Web framework trends


Gradually updating system architecture

When we started Cloudinary, we wanted to have a great image management solution available as quickly as possible. It was critical not to re-implement the wheel, to use as many existing solutions as possible, and to use powerful cloud service solutions such as AWS.

The beta version of the service was launched with quite a basic architecture. A single virtual server hosted on EC2, a Ruby on Rails web application, a MySQL database and Amazon S3 cloud-storage, and an Nginx web server with direct access to S3 buckets that contain previously uploaded and generated images. We started with Amazon CloudFront as the CDN and cache layer.

While more and more customers joined the services and migrated their images to Cloudinary, the required scale of computation power was growing rapidly. We gradually enhanced the architecture while running, as you can see in the presentation linked above.

Today Cloudinary handles images on quite a huge scale. The service deployment involves many strong EC2 server instances and the automatic scaling of the number of servers is used to handle bursts and changes in service load. Multiple AWS load balancers (ELB) feed dozens of high performance servers and the data records are stored both on MySQL sharded databases and Cassandra NoSQL clusters. The applicative layer is based on Ruby on Rails, Java and Scala modules, leveraging dozens of open source tools. Optimized image and video is delivered via the premium CDN of Akamai. We also integrated with AWS SQS for background processing queues, ElasticSearch indexing cluster for advanced search capabilities and plenty of third party image processing solutions as add-ons to the service.

Cloud service architecture

Recently we launched a powerful video management solution that joined the image management service. Supporting real-time video transcoding, resizing, manipulation and optimization required further architectural changes and enhancements that are not yet included in the diagram above.

Some internal numbers

So 3.5 years since the soft launch of our service, we have reached quite impressive numbers:

  • More than 60,000 developers have signed up to the service.
  • We have more than 2,000 paying customers from 65 different countries, ranging from startups to Fortune 500 companies.
  • About 20 Million images are added to Cloudinary's service every day.
  • About 1.5 Billion images are delivered every day via Cloudinary. If you include Cloudinary's big customers that have their own CDN layer, the actual number of Cloudinary images delivered every day is probably 100 times larger.
  • We have about 1,000 CPUs to handle all image upload and manipulation requests.

Summary

The journey so far was exciting. The bootstrap approach and the organic growth of the service and the company worked very well for us. As developers, building a service that makes the lives of many developers easier, makes us happy and proud. The scale of the service and the numbers mentioned above are quite impressive, while many web sites and mobile applications worldwide are already using Cloudinary.

The bootstrap approach made us grow the service while running and become very customer focused. It forced us to only do tasks that we were sure are important. It forced us to find creative solutions, which is great. We've been expanding the team while growing, with veterans that are absolutely amazing. Now the company is financially strong and growing rapidly with significant budget to help us reach the next level, without losing the strong effectiveness of a young bootstrapped company.

But it's just the beginning. The system architecture keeps evolving and advancing while we've just recently added video support (in addition to images) and while more and more sites and apps start using Cloudinary to manage their media assets. We have a very long list of powerful features and capabilities in our road-map, and as usual, many of them are customer requests. And still not all developers have heard about Cloudinary yet... So a lot of work is ahead of us to grow the team, the service, the architecture and the business.

If you have any questions or feedback, please send them over!

Recent Blog Posts

Build a WhatsApp Clone with Automatic Image Optimization

In the previous post, we showed how to upload images to a Cloudinary server. In this part, we will play with some of the features we see on the WhatsApp technology. After you or your users have uploaded image assets to Cloudinary, you can deliver them via dynamic URLs. You can include instructions in your dynamic URLs that tell Cloudinary to manipulate your assets using a set of transformation parameters. All image manipulations and image optimizations are performed automatically in the cloud and your transformed assets are automatically optimized before they are routed through a fast CDN to the end user for an optimal user experience. For example, you can resize and crop, add overlays, blur or pixelate faces, apply a variety of special effects and filters, and apply settings to optimize your images and to deliver them responsively.

Read more
With automatic video subtitles, silence speaks volumes

The last time you scrolled through the feed on your favorite social site, chances are that some videos caught your attention, and chances are, they were playing silently.

On the other hand, what was your reaction the last time you opened a web page and a video unexpectedly began playing with sound? If you are anything like me, the first thing you did was to quickly hunt for the fastest way to pause the video, mute the sound, or close the page entirely, especially if you were in a public place at the time.

Read more
Impressed by WhatsApp Tech? Build WhatsApp Clone with Media Upload

With more than one billion people using WhatsApp, the platform is becoming a go-to for reliable and secure instant messaging. Having so many users means that data transfer processes must be optimized and scalable across all platforms. WhatsApp is touted for its ability to achieve significant media quality preservation when traversing the network from sender to receiver, and this is no easy feat to achieve.

Read more
New Google-powered add-on for auto video categories and tags

Due to significant growth of the web and improvements in network bandwidth, video is now a major source of information and entertainment shared over the internet. As a developer or asset manager, making corporate videos available for viewing, not to mention user-uploaded videos, means you also need a way to categorize them according to their content and make your video library searchable. Most systems end up organizing their video by metadata like the filename, or with user-generated tags (e.g., youtube). This sort of indexing method is subjective, inconsistent, time-consuming, incomplete and superficial.

Read more

iOS Developer Camp: The Dog House

By Shantini Vyas
iOS Developer Camp: The Dog House

Confession: I’m kind of addicted to hackathons. Ever since graduating from Coding Dojo earlier this year, I’ve been on the hunt for new places to expand my skills and meet new people in the tech space. iOS Developer Camp’s 10th Anniversary event bowled me over. Initially, because of its length. 48 hours? Yeesh. I had no idea that those 48 hours would change my life. But let’s first get a little backstory on my favorite topic: dogs.

Read more