Student.com Migrate Kubernetes Workloads to Spot, Saving 83% on their EC2 with Ease
Seasonality makes Kubernetes Autoscaling a Necessity for Student.com
Student.com had recently restructured to use
Kubernetes when they first started looking into using AWS Spot Instances to boost cost optimization. The nature of the business (students predominantly looking for accommodation during the months running up to new terms) meant that seasonal spikes in usage were high. This, in turn, meant that all their environments needed to be able to scale freely. Kubernetes allowed this at a container level, but what was really needed to help
optimize costs was something that would help scaling at an underlying infrastructure level. And if this could be done whilst simultaneously reducing costs then all the better!
That was when Steve Lowe, CTO at Student.com, first came across Spot. Upon seeing the promise of an 80% reduction in cost with 100% availability and no architectural changes, Steve, like many others, thought it was too good to be true. After looking into the company more and understanding the premise of the offering, Steve and his team decided to test out Spot.
Simplicity Of Setting Up Results In Rapid Migration
“We’d been talking about using Spot for a while and decided to give it a go on our staging environment. We put aside a week to do that migration. In the end, it only took a few hours” – Steve Lowe
Setting up with Spot was, to quote Steve, “ridiculously easy”. Once Spot was connected with the Student.com AWS account, the Kubernetes integration meant that migrating workloads could be done with only a few clicks. Student.com just migrated the ASG of the Kubernetes nodes to Spot using the Spot Analyzer tool and connected Elastigroup by Spot to the Kubernentes Master server. Then, the Elastigroup started to utilize spot instances in the running cluster.
This meant that Student.com could migrate workloads quickly and simply, with minimal effort from their team. “It was as easy as it gets. We simply put in the details, let it go and it did the migration for us, we didn’t have to do anything!”
With Staging Migrated, Production Soon Followed
Steve’s team were more than impressed with how Spot was simultaneously efficiently managing their containerized deployments and reducing spend through the use of
spot instances. “We spoke about leaving it running as a POC for about two weeks, but the guys got a bit keen whilst I was away in China visiting the team there. By the time I came back, they had migrated over our production estate! They were very happy and very confident. Prod migration took us less than four hours.”
Now Student.com was running its Kubernetes clusters with Spot, with the platform’s built-in autoscaling features and use of EC2 Spot resulting in an 83% reduction in costs.
What’s next for Student.com?
In the few weeks since Student.com found Spot, they have migrated all Kubernetes workers and aim to migrate the masters over soon. After this, they intend to look into using Spot
MLB as well. With this completed, “we’ll probably have about 90% of our estate running on Spot.”
“I’ve already recommended Spot to friends. Certainly, anyone who’s done the
container journey to the stage where horizontal scalability is built in, using Spot on top is just a killer combination. It’s a no-brainer!” – Steve Lowe