If you've been following our blog posts over the last year, you know that the Autodesk Cloud Platform team has been moving full steam ahead with our containers strategy. Our intent was to solve very specific problems in our environment, that we felt containers could help us solve:
- Improved Efficiency. We need to improve the overall resource utilization in our data centers to aid in cost reduction as well as fully optimizing the use of our infrastructure.
- Data Center Abstraction. With a hybrid data center strategy, we want to abstract away the underlying data center provider and enable our platform services to be seamlessly deployed and able to run across multiple providers in the same way.
- Developer Efficiency. By standardizing our CI/CD process on top of a container-based infrastructure we felt we could more rapidly deploy updates and significantly simplify the developer experience for building services.
As we started out on our mission, we quickly discovered that standing up a container infrastructure, with all the required components, in a fully automated way is, well, not an easy task. Making it even more challenging was the constant barrage of information about new emerging technologies and startup companies that were popping up everywhere, offering promises of new and exciting capabilities around containers. So, while we knew we had to remain very flexible, we also knew we had to stay the course among all the distractions to build a strong, industrial strength foundation. This is where Mesos, Marathon and DC/OS come in.
Mesos and Marathon are powerhouse resource management and scheduling infrastructures that have proven to scale to handle production workloads. In our own tests, it proved to be exactly that, highly scalable with industrial strength reliability. Its because of this, we chose Mesos as the basis for our ongoing work. Deciding to use DC/OS was critical in enabling us to move quickly, by making it straightforward to get up and running with a complete Mesos-based solution. Since we don’t have an army of DevOps engineers, having the simple DC/OS installation and thorough documentation was critical.
So, fast forward to today, you might ask if we've been able to achieve the goals we originally set out to achieve when we started on this mission. Having now run our event streaming and big data ingestion pipeline services in production on DC/OS, across 3 regions, over the last year, we've achieved the following results:
- A 66% reduction in AWS Instances
- Cost Improvements up to 57%
- An impressive 40 sec time to deploy a new build with zero downtime
- A 3 min time to stand up a new region
- 100% Uptime
- Total Resources needed: 1 DevOps Engineer
Impressed? We're pretty psyched about the results of what we've done thus far and very optimistic about how we can continue to improve going forward. We're much better at utilizing our infrastructure resources and as a result, have improved our cost efficiency. We've been able to demonstrate the ability to rapidly standup up infrastructure and deploy new builds in a rolling, zero-downtime manner. We were able to accomplish all of this, with only one engineer! Not, too bad if you ask me. We wouldn’t have been able to achieve this without DC/OS.
So where do we go from here? Well, it only gets better, as we begin work on the next phase of our cloud platform, Forge!
If you haven't already heard, with the Forge Platform, Autodesk is establishing a set of cloud services, APIs, and SDKs, so developers can quickly create the data, apps, experiences, and services that power the future of making things. Enabling an ecosystem of developers makes our mission for a more efficient, cost effective, easy to use cloud platform even more essential. We believe that further adoption of our DC/OS based infrastructure and CI/CD pipeline will allow us to more quickly move to deliver cutting edge capabilities to our customers. Simultaneously, ensuring that we continue to provide the highest level of performance and availability.
We are particularly excited that DC/OS is 100% open source and proud to have been part of the team to help make it happen. Autodesk is by no means new to open source. In our Autodesk Cloud Platform team, we have most recently open sourced our Ochopod and Ochothon tools which we use along with DC/OS to enable container orchestration and CI/CD tooling. We believe that with DC/OS being fully open source (and free!), our customers will benefit from a strong community of developers contributing new capabilities to help realize other ways to extend the technology.
We're excited to be involved and to contribute to a growing community around DC/OS. If you want to follow along on our containers journey, check back for more posts on our blogs. If you're interested in learning more about Forge, visit our developer portal or better yet, meet us at our Forge DevCon 2016 conference in San Francisco in June!
And don't forget to check out dcos.io to learn more about and contribute to DC/OS.