4 minute read

Ogury banner

Ogury, a fast-scaling digital advertising company, depends on AWS and Kubernetes to harness the personification data that fuels customers’ campaigns and ensures consumers’ privacy. During its migration to Kubernetes, Ogury struggled to get visibility into its container-related cloud spend. The company had some atypical deployment variables and relied largely on open source tooling, including Prometheus and Thanos. After turning to Kubecost, Ogury was able to realize seamless integrations across its stack and gain Kubernetes cost visibility and allocation down to the service level.

Ogury: personified advertising that values personal privacy

Ogury’s mobile-optimized advertising engine is powered by an integrated adtech stack that enables brands to reach their target audience at scale, while ensuring 100% user privacy and security. Many of the world’s biggest companies use Ogury’s solutions, including Amazon, Visa, Pepsi, Nike, Adobe, Levi’s, McDonald’s, Intel, IBM, and Fiat.

The global company utilizes the AWS Cloud and Kubernetes to run APIs that enable an unparalleled command of personification data at scale. Ogury’s contextual and semantic data is augmented with unique mobile audience data, which is future-proof, cookie-less, and ID-less for privacy protection.

Modernizing around Kubernetes also called for clearer cost transparency

Ogury has been an AWS-based business since its founding in 2014, running APIs using containers on Amazon EC2 instances and initially utilizing an in-house container orchestration system. Ogury initiated its move to Kubernetes in 2019, completing the migration over the course of two years. In early 2021, Ogury achieved 100% live and fully stateless API workloads running on Kubernetes.

However, six months into the migration process, Ogury began to encounter challenges with assigning Kubernetes costs. The company’s Kubernetes deployment included Prometheus for monitoring and alerts and Thanos for adding high availability, long-term data storage, queries, and metrics. Ogury initially tried using a homemade Grafana dashboard to capture Prometheus metrics and assign shares of API calls (and associated memory usage) to attribute costs and optimize resources. Unfortunately, this approach was overly complicated, and the experience left the team seeking a more stable, effective, and efficient path forward.

Ogury’s technical and finance teams work in tandem to forecast yearly budgets and assign costs and revenues to projects. The core infrastructure technical team is responsible for cloud management strategy, including tagging cost centers and performing cost optimizations. Post-migration, Kubernetes became the technical team’s single biggest cost tag—and visibility into where exactly those costs were coming from was poor. The company’s technical and finance team leaders agreed that finding a solution to better understand their Kubernetes-related cloud costs was a top priority.

Open to open-source: Kubecost delivers easy integrations and granular cost allocations

Ogury vetted several large providers of cloud cost visibility solutions. However, they quickly found that these solutions all required purchasing an expensive suite of full-blown cloud cost management software, whereas Ogury needed a tool purpose-built for understanding Kubernetes costs. The other providers offered cost models based upon a percentage of total cloud spend, which was not reasonable given Ogury’s use case.

In early 2020, Ogury discovered Kubecost through a blog announcement on Grafana and, after vetting, signed on as an Enterprise-tier Kubecost customer. The dedicated support quickly proved advantageous in completing integrations across Ogury’s atypical environments. Ogury was in frequent contact with Kubecost support throughout the first six months of the engagement to set up Kubernetes cost visibility across Ogury’s deployments, especially when it came to navigating the company’s relatively complex Thanos backend.

With Kubecost in place, Ogury allocates Kubernetes spending on a per-workload, per-category, and even per-service basis. The technical team leverages Kubecost’s cloud reconciliation tools to view detailed cost management reports, which include actual costs rather than just estimates based on catalog prices. This visibility is especially beneficial given Ogury’s utilization of spot instances with highly variable costs.

Kubecost equips Ogury’s technical team with an on-demand tool for achieving specific cost estimates, investigating workload traffic anomalies, and understanding the behavior of individual services. The team also leverages Kubecost to export data to a shared, pre-aggregated dashboard to visualize Kubernetes costs and build out reports for the finance team.

Results delivered: optimized Kubernetes tooling and strategic cost visibility

Kubecost has enabled Ogury to easily integrate and optimize its support tools, ingress controller, and Prometheus and Thanos deployments. It also allows the company to achieve Kubernetes cost allocation per service at any level of granularity required, down to the namespace, deployment service, or any other type of construct.

“The first thing I liked about Kubecost was its ability to integrate into any standard open source technology stack,” says Adrien de Castelnau, a Senior DevOps Engineer at Ogury. “Our ability to plug Kubecost into Thanos and not have to deploy a specific backend was a strong positive. With Kubecost, we can now look at Kubernetes costs as granularly as we need and from whatever perspective we like, giving us a tremendous advantage in strategizing around cost and quickly addressing unexpected changes.”

Scaling with Kubernetes costs continually optimized

Ogury’s Kubernetes environments now handle 10,000,000,000 API calls daily, an 8x increase over the past three years. With the company’s Kubernetes migration complete, Ogury leverages Kubecost for granular visibility into cost allocation and uses Kubecost enterprise support for assistance when implementing new product features. The company plans to introduce anomaly detection alerts to enable proactive investigations into surprise service cost increases. Ogury also plans to expand its Kubecost usage across more of its technical teams (beyond the core infrastructure team and even to the finance team) to further increase productivity and cost optimization across the organization.