International. Netflix's journey to the cloud began in August 2008, when we had a very serious problem with the database and were unable to send DVDs to our members for three days. There we realized that we had to switch from the unique vertical-scale points of failure (the relational databases in our data center, for example) to other distributed, highly reliable, and horizontally scalable cloud systems.
We chose Amazon Web Services (AWS) as our cloud provider because they offered us the highest scale and the widest range of services and features. By 2015, we had migrated most of our systems, including all customer-facing services, to the cloud. However, we wanted to take the time to determine a secure and durable cloud path for our billing infrastructure, as well as for all aspects of our customer and employee data management.
We are pleased to announce that in early January 2016, after seven years of effort, we completed the migration to the cloud. And we already turned off the last bits of the only data center that was still using our streaming service.
The move to the cloud has brought many advantages for Netflix. The number of broadcast members we had in 2008 has increased eightfold. In addition, they are much more active, as evidenced by the total number of views, which tripled in the last eight years.
Netflix's own product has continued to evolve rapidly, incorporating many new resource-hungry features, and relying on ever-increasing volumes of data. It would have been very difficult for us to sustain such a rapid growth of our data centers; we wouldn't have had time to even mount the servers on the shelves.
The elasticity of the cloud allows us to add thousands of virtual servers and petabytes (millions of gigabytes) of storage space in a few minutes, and that's why this expansion has been possible. On January 6, 2016, Netflix expanded its service to more than 130 new countries, becoming a truly global Internet television network.
Being able to use multiple AWS Cloud Regions, spread around the world, allows us to dynamically switch sites and expand the capacity of our global infrastructure. This way we can create a better streaming experience for Netflix members, regardless of where they are.
We rely on the cloud for everything we need in terms of scalable computing and storage space: our business logic, distributed databases, and big data processing and analysis, recommendations, transcoding, and hundreds of other features that make up the Netflix app. The video is streamed through Netflix Open Connect, our globally distributed content delivery network to efficiently send our bits to members' devices.
The cloud has also allowed us to significantly increase the availability of our service. In our data centers we sometimes suffered from blackouts, and while we had some inevitable difficult times in the cloud, especially the early days of migration, we have seen a steady increase in our total availability, getting closer and closer to our goal of four nines (99.99%) of service uptime.
It is impossible to avoid failures in large-scale distributed systems, even if they are cloud-based. However, the cloud makes it possible to generate highly reliable services from fundamentally unreliable but redundant components. By incorporating the principles of redundancy and degradation correctly into our architecture, and by maintaining the discipline of regular production drills with Simian Army, we can survive failures in cloud infrastructure and within our own systems without impacting the member experience.
Cost reduction was not the main reason for our decision to move to the cloud. But it is undeniable that our cloud costs per transmission start are much lower than those of the data center, a more than welcome side effect. This is possible thanks to the elasticity of the cloud, which allows us to constantly optimize the mix of instance types, and grow and decrease the space we occupy almost instantly without having to maintain large capacity buffers. We can also take advantage of the economies of scale that can only exist in the large cloud ecosystem.
Given the obvious advantages of the cloud, why did it take us seven whole years to finish the migration? The truth is that the move to the cloud was a lot of work, and we had to make very difficult decisions along the way. Surely the easiest way to move to the cloud would have been to take all the systems, without changing them, out of the data center and release them into AWS.
But if we had done that, we would have taken with us all the problems and limitations of the data center. Instead, we opted for the "cloud-native" approach: we virtually rebuilt all of our technology and radically changed the way we run the business. From an architectural standpoint, we migrated from one monolithic application to hundreds of microservices, and denormalized our data model using NoSQL databases.
Budget approval, centralized version coordination, and multi-week hardware provisioning cycles gave way to continuous delivery and independent decision-making teams with self-service tools in a direct-attached DevOps environment, helping to accelerate innovation.
We had to set up many new systems as well as learn new skills. Transforming Netflix into a cloud-native company took a lot of time and effort, but it put us in a much better position to continue to grow and become a global TV network.
Netflix's streaming technology has come a long way in recent years and it's nice not to feel the limitations we had before. Since the cloud is something fairly new to many in our industry, there are still many concerns to be answered and many problems to solve. Through initiatives such as Netflix Open Source, we hope to continue collaborating with very bright technological minds and face all these challenges together.
Text written by Yury Izrailevsky, Vice President, Cloud and Platform Engineering at Netflix.
Leave your comment