Transforming the education sector, one test at a time

Open Assessment Technologies collaborates with world-class educational institutions to drive digital transformation via the TAO assessment tool. Open Assessment Technologies has administered over 100 million exams in more than 30 languages.

To meet the unique requirements of its user base, the company runs its assessment application in the cloud. “Kubernetes has been part of our journey since the beginning,” said Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies.

Managing infrastructure manually was time-consuming

TAO is a product that experiences dynamic changes in demand. The number of users may rise sharply during campaigns; at other times, there may be very few of them using the platform. The dramatic changes in usage patterns required the operations team to manually scale cloud capacity up and down in line with the forecasted demand. This approach quickly turned out to be unsustainable. 

We know that next week we will have a campaign, and that this campaign will involve a certain number of users, even thousands of them. What I did was prescale my infrastructure to have that amount of resources available for the users coming in. But doing this manually took too much of my time.

Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies

The team initially started looking for a cost visibility solution that would display a breakdown of expenses by cluster. “It’s very hard to find that information using Google Cloud Platform. You need to input queries and stuff like that. I needed to find something to give me a report on how much I spend on my cluster,” said Rafael Tovar.

Looking for a solution: Kubecost vs. CAST AI

Initially, the team tested Opencost, the open-source tool from Kubecost. 

Kubecost has an open-source tool, but the installation and handling of the infrastructure were time-consuming. So, I just started looking and found CAST AI. The solution provided the report I needed for free. Kubecost doesn’t provide data on how much you spend per cluster for free. I finally could see how much I paid for CPUs, memory, and the overall application.

Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies

Implementing CAST AI was a straightforward process, where Open Assessment Technologies was supported at every step. “Onboarding was super easy because the installation process was very smooth,” added Rafael Tovar.

I am super happy with the support, to be honest. I have a private Slack channel with the CAST AI team. After sending a message, I get a fast response. Sometimes the team needs time, but they’re always willing to schedule calls, and I love that. It’s a really lovely way to fix problems.

Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies

Slashing cloud costs by 35%

After implementing CAST AI, the team used multiple automation features to drastically reduce cloud waste and is on its way to achieving a spectacular 35% cloud cost savings.

We expect to reach savings of around 35% next year due to the fact that we have previously purchased CUDs with Google. They will end next year, and we’re bound to see more savings then.

Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies

“The biggest advantages of CAST AI for us are its ability to scale workloads up and down, pick the most optimal cloud resources for our requirements, and cost-optimize our applications automatically,” he added.

Two CAST AI features that made the difference

Node templates

CAST AI automatically analyzes the requirements of an application and provisions cloud resources that match application demand without the risk of overprovisioning or performance degradation. Node templates allow users to define virtual buckets of constraints for this decision-making process.

Without CAST, I need to manually define a certain amount of resources: CPUs and memory. Unfortunately, Google Cloud Platform doesn’t provide this customization in brief details. Defining them results in a gap of around 50-60%, which is a waste of resources. 

When I introduced CAST AI, I noticed that when I created my pods, the newly created infrastructure was based on the exact requirements I needed to meet. So, there was no more utilization gap.

Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies

Instant rebalancing

Rebalancing is a CAST AI feature that allows a cluster to reach the most optimal and up-to-date state with just one click. During the process, CAST AI automatically replaces suboptimal nodes with new ones, which are more cost-efficient and run the most up-to-date node configuration settings. 

This feature turned out to be immensely helpful to Open Assessment Technologies. “We run different applications and scale them down during the night. Once done, we use CAST AI’s rebalancing to save money in clusters across all of our environments,” said Rafael Tovar.

Our clients carry out campaigns, which is the challenge here. We need to ensure the infrastructure meets specific requirements around CPU and memory. We have maybe ten thousand users coming in when a campaign starts, and on other days, the system is not being used at all. We found CAST AI to be a perfect match for all these cases. Our production, staging, and development environments are handled 100% by CAST AI.

Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies

Massive time savings and trust in automation 

CAST AI’s automation features generated significant cost savings for Open Assessment Technologies, but also freed up the company’s Google Cloud Operation Leader to dedicate his time to more impactful tasks.

If you’re an administrator in a Kubernetes environment and you don’t have this kind of tool, I don’t know how you sleep at night. Thanks to CAST AI, it was the first time that I took a vacation for a month, and nobody asked me to add more nodes to their applications because they’re running a new campaign. I was super happy because of that.

Rafael Tovar, Google Cloud Operation Leader at Open Assessment Technologies