Building resilient futures: the 2024 State of Production Kubernetes
Exclusive new research into the trends shaping enterprise cloud native today.
Snackpass is a software platform for quick-service restaurants (QSRs). From its early days as a mobile ordering app for students at Yale university, it has reimagined the QSR technology stack from the ground up to be modern, elegant and spark joy at every customer touchpoint, from point of sale to loyalty.
Learn more at snackpass.co
In its early days, Snackpass had deployed its monolithic backend services in Heroku. But as the company’s growth accelerated, Principal Engineer Nathan Probst knew that something needed to change.
“We needed to ensure stability, gain access control capabilities and guarantee cost savings compared to the previous setup”, said Probst. “As we added more senior people to the team, we also wanted to have the option of breaking up our monolithic codebase into more fine-grained services”.
Drawing on their previous experience, Probst’s team settled on Kubernetes, specifically AWS’s managed service, EKS. EKS would allow Snackpass to scale quickly and expand into multiple regions to support high availability architectures.
While EKS provided the infrastructure, Probst knew that Kubernetes was complex, and his team needed the right management platform to drive it and support it effectively. In Spectro Cloud Palette, he found exactly what he was looking for.
The Snackpass team had discovered Spectro Cloud through its support for Kairos, an OSS project that builds immutable Linux OS images for secure edge Kubernetes. “I'm an open source enthusiast,” said Probst.
“Any company that helps put these projects out in the world stands out in my evaluation”
After thorough evaluation, Palette stood out for its simplicity of operation, trusted technical support and flexibility, including the ability for Snackpass to bring its own tools.
The Palette platform enables Snackpass’s platform engineers the ability to set up and manage their EKS clusters with minimal effort and optimize their cloud costs — backed by a support team they could trust.
“All of the interactions with Spectro Cloud’s team just continued to build comfort and confidence. It was a night and day difference to experience this depth of technical expertise compared to where we were coming from” said Probst
Snackpass used Spectro Cloud Palette to create new EKS clusters in us-east-1, incorporating additional software integrations such as Cluster Autoscaler and ElastiCache. These are all defined in what Palette calls ‘Cluster Profiles’, a blueprint of the desired state of the cluster used for declarative management.
The new EKS architecture includes dedicated clusters for development and testing purposes, with production clusters created in advance, for multiple testing cutover steps before the planned go-live date.
Snackpass decided to use its existing Mongo Atlas database to minimize downtime during cutover. DataDog provides observability for the infrastructure and application.
The migration process itself was a success: fast, smooth, with a zero downtime application cutover.
"Spectro Cloud gave us that sweet spot between Heroku and AWS" said Probst. "We have the power of Kubernetes without all of the headaches.”
A few months after being in production with Palette, Snackpass set up a cross-region failover environment, which was a key strategy to ensure reliability of its service for customers — especially as Snackpass grows to support bigger brands with hundreds or thousands of restaurants.
“Using Palette to set up the failover environment was just as easy as the first time,” said Probst.
Snackpass has big ambitions to scale, transforming the software stack for QSR operators large and small. A big part of that is evolving its infrastructure architecture to support greater scale and resiliency.
While today Snackpass’s QSR platform runs in the cloud, it has a vision to deploy it directly in restaurants, avoiding the risks caused by connectivity outages.
“Spectro’s edge offering really tipped the scales for us when we were first evaluating what it could do,” said Probst. “The ability to run on-prem Kubernetes with high availability for high-value restaurants is key for us, and we’re looking forward to getting started with Palette at the edge as soon as possible.”
By managing EKS clusters with Palette, Snackpass has realized an approximate 50% cost savings over what it had been spending with Heroku.
With observability and autoscaling, Snackpass can efficiently manage variable application demand, and grow its business while optimizing cost.
Spectro Cloud’s support team acts as an extension of Snackpass’s team’s expertise at every stage, from configuring autoscalers to troubleshooting issues.
Connect with us
.webp)
.svg){kind=logo}
