Company
Flowcore is the startup behind a developer-first platform Flowcore Streams which offers flexible, ready-to-go Kafka clusters. Revolutionizing the world of data, Flowcore’s solution allows teams to implement real-time data streaming without the complex, time-consuming setup and management.
Challenge
As Flowcore scaled its platform and onboarded new customers, the rise in data volumes made the challenge of cost control more pressing. The company started looking for a solution that could both improve cost visibility and optimize costs in a dynamic, high-traffic environment to ensure efficient scaling without overspending.
Solution
Flowcore tested several solutions including OpenCost, but only Cast AI delivered the level of cost visibility and automated savings. Instead of constantly monitoring and adjusting resource limits manually, Flowcore can rely on Cast AI to provide an automated and efficient solution for cost management.
Results
- 50% cloud cost savings
- Dramatic improvement in resource allocation efficiency
- In-depth cost visibility across every cluster
Node rebalancing
Cast AI maintains optimal compute cost by running a nightly scheduled rebalancing job. During rebalancing, Cast AI bin-packs workloads into fewer nodes and then moves them to more cost-effective compute resources.
Cost savings achieved via the nightly rebalancing
Cost reduction via rebalancing
This is how Cast AI helps Flowcore reduce cloud waste at the node level, resulting in optimal resource utilization and compute cost savings.
Dramatic resource utilization improvements
Workload autoscaling
Cast AI workload autoscaler unlocked additional saving by rightsizing down both CPU and Memory requests at the workload level.
With Cast AI, we can keep our cloud expenses at around 50% of the original costs. Additionally, we noticed another impactful use case – not just cost savings, but also the efficiency of the cluster itself.
We had a cluster that was experiencing issues with nodes going down, and while services seemed to be running fine, our team had to deal with the heaviest services running in a degraded state.
After enabling Cast AI and adding workload autoscaling, we saw significant improvements. Although the savings on that cluster were only around 5-10%, everything was properly allocated, meaning no nodes or workloads were starved for resources.
Julius á Rógvi Biskopstø
CTO and Co-founder at Flowcore
Controlling a spike in cloud costs while scaling the product
When did the rising cloud expenses become an issue for your company?
When we started to scale our platform and onboarded new customers – especially larger ones with a massive amount of data flowing through our platform – cloud costs quickly became an issue.
Initially, I used tools like Kubecost to monitor the cost of things. Then I came across Cast AI, and enabled it in read-only mode almost right from the start, just to see how much everything was costing.
At the time, we were on an Azure Founders Hub subscription, which gave us a lot of credits to keep everything running. As soon as we reached the end of the 150K credits and costs started coming in, that’s when we fully enabled Cast AI. Since we had already gathered all the data, we were able to get up and running quickly with the correct settings in place.
Why did you look beyond Kubecost with cost control in mind?
For me, the key difference between Kubecost and Cast AI is how they handle costs. With Kubecost, you dive straight into workload-level costs, which is useful, but with Cast AI, you get visibility into both the workloads and the underlying infrastructure – like machine and memory costs.
This dual-layer visibility in Cast AI is a significant advantage because it gives you a more complete overview of everything, not just the workloads. This is a very important difference when you’re managing costs across the entire environment.
How quickly did Cast AI generate cost savings for you?
That was basically immediate. We had Cast AI running for several months in read-only mode before we activated it. I think it’s a great feature to have for free – being able to connect your cluster to Cast AI and see exactly what you’re missing. Then, once you activate it, you get all the benefits right away.
Reducing cloud costs by 50% with automation
What results did you achieve with Cast AI?
With Cast AI, we can keep our cloud expenses at around 50% of the original costs. Additionally, we noticed another impactful use case – not just cost savings, but also the efficiency of the cluster itself.
We had a cluster that was experiencing issues with nodes going down, and while services seemed to be running fine, our team had to deal with the heaviest services running in a degraded state
After enabling Cast AI and adding workload autoscaling, we saw significant improvements. Although the savings on that cluster were only around 5-10%, everything was properly allocated, meaning no nodes or workloads were starved for resources.
We didn’t experience the “noisy neighbor” effect anymore, where resources are quietly being starved without visible errors, leading to strange behavior like nodes shutting down unexpectedly.
With Cast AI, we could see everything running more smoothly, and network throughput improved significantly. Services had the overhead they needed to spin up new threads and handle workloads properly. So, it’s not just about reducing costs, but also about getting more value and performance for each dollar spent, which is crucial.
Which Cast AI features had the biggest impact on your cloud bill?
We’re using everything: workload autoscaler, node autoscaling, node rebalancing, and even horizontal autoscaling on some services.
The biggest impact so far has been from the node rebalancer. We’re running on Azure, which has limitations on the number of machines and CPU quotas in regions, it’s a bit restrictive. But by specifying the right requirements and ensuring we stay within those quotas, the rebalancer has delivered significant savings by using the correct CPUs.
Your reports have been excellent, especially during the onboarding process. We received tips on nodes that were running inefficiently and got guidance on how to manually rebalance them to save even more before the automated compaction took place.
What was the support you received from the Cast AI team like?
It was a very active onboarding from your side, with observers on the cluster to ensure everything was optimized, and that really helped a lot.
One thing I’d add is the dedicated Slack channel you get is extremely valuable. It’s great how you can quickly add people to it if needed, and the response time on issues is very fast. Usually, problems get resolved right away, and if you need information, you get it quickly.
The level of activity and responsiveness in the channel set up for us by the Cast AI team is really unique compared to the slower, more traditional support ticket systems. It makes a big difference.
The impact of automated cost optimization
What are the biggest benefits of Cast AI for Flowcore?
The biggest benefit is having control over our costs. You can see exactly what everything costs and manage it accordingly. Another key aspect is efficiency – you can be sure that the services you’re running have enough breathing room and are more stable overall.
The dollar amount saved is important, but just having visibility and control over it is a huge deal. Without that, it can feel like a black box. Even though we use other observability tools and APMs, having transparent cost visibility makes everything much clearer.
What was the impact of the achieved cloud cost savings for your company?
We’re a startup, and we’re still in the early stages, so saving a few thousand dollars a month gives us some much-needed breathing room. It offers peace of mind because we don’t have to immediately focus on generating revenue streams just to cover those costs.
Instead, we can concentrate more on building our core functionality without feeling pressured to roll everything out at such an extreme pace. It really provides that extra breathing room we need.
Bringing Cast AI to Flowcore’s customers
How do the benefits of Cast AI translate into your customer relationships?
We have some customers whose Kubernetes setups we manage, and we install Cast AI for them to gain observability into the infrastructure and costs. This allows us to show them how much they’re spending and how they could potentially save.
For every cluster we deploy, especially when we’re setting up a dedicated installation of our platform, we’ll add Cast AI to optimize costs for the customer automatically.
For one of our customers, we installed Cast AI right away to get cost visibility and show them how much managing it would cost. Once they were ready, we could just turn it on since the data was already in place. It was a smooth process.
It’s always great to show our customers that we can help them save money. Essentially, we’re helping them recoup the costs they’re spending on us. It’s almost like offering them a free subscription, but we’re achieving that through the use of tools.
