The existing compute model relied heavily on fixed resources, driving up costs without matching utilization. This was unsustainable as traffic patterns fluctuated across global time zones.
The absence of automated scaling and reliance on traditional hosting methods created risks during seasonal or promotional spikes. Outages during these periods had a direct impact on order fulfilment and customer satisfaction.
Manual and error-prone deployments slowed down release cycles, delaying feature rollouts and reducing the ability to respond quickly to market demands.
Without unified monitoring, logging, or APM integration, troubleshooting was reactive and time-consuming, leading to extended downtimes and delayed incident response.
Managing access for a growing engineering team across Kubernetes clusters was complex, lacked single sign-on, and increased operational overhead.
Kubernetes Modernization: Set up a self-hosted Kubernetes cluster optimized with spot instances for cost efficiency, supported by a NetApp storage solution for reliability.
Deployment Automation: Established GitOps workflows using ArgoCD, CircleCI ensuring faster and error-free deployments.
Security & Access Control: Implemented Single Sign-On (SSO) with role-based access to streamline user management across namespaces and resources.
Observability & Monitoring: Integrated Datadog and New Relic APM for end-to-end monitoring, real-time issue tracing, and application performance insights.
Centralized Logging: Deployed Elasticsearch to create a unified and searchable logging solution for all applications.
(self-hosted, spot instance optimization)
Improved Search Reliability – Solr cloud cluster scaled effortlessly to handle peak holiday traffic.
Operational Excellence – real-time monitoring reduced mean-time-to-resolution (MTTR).
Faster Delivery – GitOps-driven pipelines accelerated deployments.
High Availability – resilient infrastructure ensured uptime during peak traffic.
Secure Access – simplified user management with enterprise-grade SSO integration.
