Explore ThreatMark’s AI-based solution for real-time fraud detection. Dive into their hybrid setup with EKS, NLB, Karpenter & RKE2 for optimal performance.
At a Glance
Challanges
- Enhanced security (Banking industry)
- Cloud and on-premise solution
- Ultra low latency
Benefits
- Highly available and scalable Kubernetes cluster
- Fully automated infrastructure (both cloud and on premise)
- Low latency solution
- Focus on security
Objectives
ThreatMark offers Fraud Prevention Solution, based on AI, to its customers that consists of Threat Detection, Identity Verification and Transaction Risk Analysis.
ThreatMark needed a scalable Kubernetes service running on hybrid solution with an enhanced security. The lowest latency possible was required as this was crucial for “real-time” detection and analysis.
Solution
The solution leverages Amazon Elastic Kubernetes Service (EKS) with Network Load Balancer (NLB) and Karpenter for auto-scaling in the cloud environment. EKS provides a managed Kubernetes service with robust scalability and reliability, while NLB enables efficient load balancing for incoming traffic and low latency. Karpenter, an open-source Kubernetes scaling engine, is used for dynamic pod scaling based on resource utilisation, workload demands, and custom policies, ensuring optimal resource allocation and cost efficiency. Calico was chosen as the main network plugin because of its low latency networking.
For the on-premise solution, the solution utilises RKE2, a superficial Kubernetes distribution designed for edge computing and on-premise deployments.
Petr Joachim
VP OF ENGINEERING
We found DevOpsGroup when we needed to help setting up custom Kubernetes cluster setup on AWS for our product. It has many caveats and specific requirements regarding security and costs and guys from DevOpsGroup really nailed it. They were able to quickly adopt to our working schedule and they started to be productive quickly. They are very independent professionals with baked in project management in the team so it worked smoothly
