BliBli is a rapidly growing e-commerce site developed specifically to encourage online shopping in Indonesia. It has grown rapidly since its founding in 2011 by offering products and services targeted to its local markets and investing in technologies that deliver a superior customer experience.
Continually improving the customer experience is among the highest priorities for the company’s IT department. “Our main focus is customer satisfaction,” says William Chewardy, principal lead performance engineer at BliBli. “It is essential that we offer all our customers a secure, efficient, and highly satisfying shopping experience. That means delivering a stable, high-performance infrastructure that can scale at the pace the business grows.”
To meet that objective, BliBli has been adopting modern IT best practices and technologies. It adopted a container approach, using Docker and Kubernetes, and shifted its entire IT infrastructure to Google Cloud Platform (GCP).
Observability a key requirement for microservices
The adoption of new IT practices presented a number of challenges, Chewardy says. “In 2015, we moved from a monolithic architecture to one based on microservices, and today we have around 350 microservices, which creates a complex environment to manage. To support so many services, observability is critical for us. If incidents occur that interfere with our website performance or transaction systems, we need to be able to identify the root cause quickly. Any change, for better or worse, in the customer experience will show up in our conversion rates, so the metrics we track on application performance are very critical to us.”
BliBli’s solution for observability is New Relic. The company employs the full range of products that comprise the New Relic One platform to provide observability across all systems, including the development environment, customer-facing websites, and e-commerce applications. Because New Relic One unifies all telemetry data in a single platform, it gives BliBli a comprehensive yet easy-to-understand picture of everything happening throughout its complex ecosystem of applications and services. Specifically, it proactively pinpoints issues before they become problems, by quickly identifying relevant relationships among growing collections of observability data.
Prior to adopting New Relic, BliBli used Prometheus for event monitoring and alerting, but it was neither as comprehensive as New Relic in the breadth of data it collected nor as easy to use.
The impact of New Relic has positively affected multiple operational areas within BliBli, Chewardy says.
In software development, New Relic addresses one of IT management’s chief concerns: deploying new features and functions in the timeliest manner to enhance the customer experience. “The pace of software deployment makes it challenging to track the performance of all applications, understand if they are performing as expected, and deliver the quality experience we want our customers to have,” says Chewardy. “The seamless integration of New Relic with our Docker and Kubernetes environment provides valuable, detailed information that makes our software development more efficient and effective.”
As an example, he said, “for a major deployment involving many services, it’s difficult to verify the application behavior post-deployment. It can take hours to simulate different loads and analyze the results. With New Relic, we can record and monitor deployments easily. Using deployment markers, it’s easy to correlate the deployment to critical performance metrics. We can verify deployment performance in under one hour.”
During the migration from on-premises infrastructure to GCP, BliBli used New Relic to perform load testing for each service as it was migrated, to identify any degradation in service availability or performance. “New Relic’s detailed performance data makes it easier for us to analyze the migration process and show progress,” Chewardy says. Post-migration, BliBli uses New Relic to optimize cloud resource usage and control costs. Chewardy indicates that after two months, costs were reduced by about 30%.