Ctrip International Ticketing Cloud‑Native Migration: Infrastructure as Code, Logging, Monitoring, and Cost Optimization

Background – To support overseas users, Ctrip International Ticketing sources data from global suppliers and runs services in many regions. Public cloud was chosen over building private data centers for flexibility and cost.

Cloud‑Native Adoption – The team follows established cloud‑native standards to build scalable, highly available, loosely coupled applications, focusing on rapid, low‑cost service delivery.

2.1 Infrastructure as Code – All infrastructure is defined in version‑controlled IaC repositories alongside application code. Terraform is used for declarative provisioning of managed Kubernetes clusters and other resources, enabling reproducible environments and CI/CD integration.

2.2 Logging – A managed Elastic Search service is used for log storage. Logs are collected via a DaemonSet on each node, forwarded to stdout/stderr by applications, and then processed and visualized in Kibana, decoupling logging from business code.

2.3 Monitoring – The monitoring stack consists of Prometheus + Grafana, deployed with the Prometheus Operator and integrated with Thanos for high‑availability, long‑term storage, and multi‑cluster aggregation. Thanos Sidecar uploads data to S3, while Thanos Compact downsamples old data to reduce storage costs.

3.1 Compute Cost Optimization – Elastic scaling is achieved with Kubernetes HPA and Cluster Autoscaler, automatically adjusting pod replicas and node counts based on load. Spot (bid) instances are mixed with on‑demand instances, using node affinity to run tolerant workloads on cheaper spot capacity while keeping critical services on stable nodes.

3.2 Storage Cost Optimization – Historical log and monitoring data are periodically snapshotted and moved to low‑cost object storage. Serverless functions (e.g., AWS Lambda) run lightweight backup scripts, charging only for execution time.

3.3 Network Cost Optimization – For outbound‑heavy ticket queries, a transparent Squid proxy is deployed in a private subnet to route external traffic, allowing the use of outbound‑only pricing models.

Conclusion – By adopting cloud‑native practices, Ctrip International Ticketing built a stable, automated production environment that accelerates delivery, improves elasticity, and reduces operational costs, while enabling rapid feedback through centralized logging and monitoring.