Use case
Microservices Deployment on
On-Premises Environment
- Dhruv Bundheliya
About the Customer
Neuralogitec is based on visual Deep-AI algorithms and big data, adding a new dimension of component visual analytics to the SMT machine metadata created in the production line. The NeuraLogic Deep Visual-AI platform aggregates and analyzes images and data from 100% of the electronic components. It combines existing production data with unique visual data collection to ensure product quality, authenticity, and traceability, specifically for OEMs and EMSs.
Customer Challenge
Neuralogitec faced two critical challenges: achieving real-time responsiveness and optimizing cloud infrastructure costs. Their existing cloud-based solution struggled to meet the required response times for image predictions, which affected the efficiency of their AI/ML models.
The key technical requirement was to enable low-latency predictions and maintain high availability while minimizing the infrastructure costs. Additionally, Neuralogitec needed a secure Kubernetes cluster with Role-Based Access Control (RBAC), network policies, and an automated CI/CD pipeline for deploying containerized AI/ML models. Without addressing these challenges, the prediction response times would remain high, impacting the performance of the models and raising operational costs.
Solution
IAMOPS proposed a solution that involved migrating from a cloud-based infrastructure to an on-premise deployment. This allowed Neuralogitec to have better control over the latency and improve real-time prediction performance. The solution leveraged Kubernetes on Ubuntu servers and included the following elements:
- High Availability for K3S API Server: HA-Proxy was configured as a load balancer to distribute traffic to the Kubernetes API servers, eliminating the risk of a single point of failure.
- Automatic Failover: To ensure uptime, Keepalived was implemented on the master node. In case of a node failure, automatic failover was set up to keep the Kubernetes API available, minimizing disruptions.
- Health Monitoring and Notification: A health check script was integrated into HA-Proxy, which monitors the API servers. In case of failure, notifications were sent through Microsoft Teams to alert the operations team, ensuring a quick response.
The solution also incorporated Kubernetes security best practices, such as RBAC for access control and mTLS authentication for the Kubernetes API server. The network was segmented with policies to ensure restricted communication between the different services, enhancing security.
Master/Slave architecture
Results & Benefits
The migration to an on-premise Kubernetes setup resulted in significant performance improvements for Neuralogitec’s AI/ML platform. Response times for real-time image predictions dropped to under 32ms, allowing the AI models to operate at full capacity.
By migrating on-premise, Neuralogitec was able to streamline their cloud infrastructure costs while achieving their target for low-latency predictions. This solution enhanced operational efficiency by reducing infrastructure overhead and maintaining high availability with automatic failover mechanisms.
About IAMOPS
IAMOPS is a full DevOps suite company that supports technology companies to achieve intense production readiness.
Our mission is to ensure that our clients’ infrastructure and CI/CD pipelines are scalable, mitigate failure points, optimize performance, ensure uptime, and minimize costs.
Our DevOps suite includes DevOps Core, NOC 24/7, FinOps, QA Automation, and DevSecOps to accelerate overall exponential growth.
As an AWS Advanced Tier Partner and Reseller, we focus on two key pillars: Professionalism by adhering to best practices and utilizing advanced technologies, and Customer Experience with responsiveness, availability, clear project management, and transparency to provide an exceptional experience for our clients.
