MakerBot migrates infrastructure from AWS to GCP with SADA

One of the largest desktop 3D printer manufacturers, MakerBot, believes 3D printing should be accessible and affordable. Along with a robust 3D ecosystem of hardware, software, apps, support, and partnerships, it runs the largest 3D printing community in the world – Thingiverse – with more than 3.5 million registered users who share 1.8 million 3D printable designs.

The Thingiverse community logs on to Thingiverse to download or share user-created design files.  Users of Thingiverse include students, educators, designers, 3D printing enthusiasts, and professionals.  MakerBot customers can also take those files and prepare and print them on their 3D Printers using MakerBot CloudPrint^TM  which enables a seamless cloud-connected experience from content to physical part.  

Challenge

To better support the Thingiverse community, which downloads 20 million files a month, MakerBot’s DevOps and Web teams decided that migrating from Amazon Web Services (AWS) to Google Cloud Platform (GCP) would significantly reduce expenses, increase scalability, and boost performance to deliver a better user experience.

After spending about three months trying to figure out how to migrate its massive database, the MakerBot team realized they needed to pull in some help due to the complexity and sheer size of the project. “We didn’t know how to do this migration right, and we wanted to make sure that it was executed with minimal downtime,” said Erik Ahrend, Lead Cloud Architect. “If Thingiverse goes down, people can’t log in to the community and download needed files”.

The migration would be a high-stakes project—the team projected that 24 to 36 hours of downtime would be needed for the entire site to copy everything over and bring it back up again. “If that process failed, it would be hours wasted and we would have to try again,” said Ahrend.

Solution

SADA went all-in with MakerBot, planning, executing and completing the massive project in eight weeks, with downtime limited to just four hours.

The team also moved MakerBot’s hosted Kubernetes cluster from AWS’ EC2 to Google Kubernetes Engine which eliminated the company’s scalability issues. “Google’s managed Kubernetes cluster has been fantastic. Simply stated, the networking part of Google is a lot simpler to understand than AWS, and it is a lot less expensive,” said Ahrend. “We are using it as our main deployment method, our container orchestration service, as well as our main web service infrastructure. It removes the pain of actually hosting everything and eliminates previous scalability issues.”

Results

  SADA enabled MakerBot to:

• Migrate underlying databases supporting thingiverse.com to GCP in a timeframe that far exceeded MakerBot’s downtime goals
• Achieve 30% cost savings per month compared to AWS
• Serve more customers with less latency 
• Ensure that GKE cluster autoscale settings are optimized
• Migrate logging away from Sumo Logic to Google Stackdriver
• Leverage Google Cloud managed services and NoOps services
• Improve network performance across a globally distributed user/customer base