Yahoo Japan Consolidates 164 OpenStack Clusters into One

What Happened

Yahoo Japan's parent company, LY Corporation, is undergoing a significant transformation by consolidating its 164 OpenStack clusters into a single cloud infrastructure named "Flava." This decision comes as the company aims to streamline operations and enhance the reliability of its services, which cater to around 300 million monthly users.

The Issue

Previously, Yahoo Japan's cloud infrastructure was heavily customized, making upgrades challenging and complicating maintenance. According to Ryuutarou Inoue, head of LY’s Cloud Infrastructure Unit, the legacy system's complexity hindered the ability to implement timely updates and security patches. The new strategy focuses on adopting a more conventional version of OpenStack, minimizing custom modifications to facilitate easier upgrades.

A New Approach

The new cloud architecture, Flava, will operate on a much larger scale with 500 hosts and over 9,000 virtual machines (VMs). This design aims to achieve three key objectives:

1. Pursuing Statelessness: By defining VM root disks as temporary, persistent data is moved to external storage, reducing service disruption during failures.
2. Application-Driven Availability: Instead of relying solely on infrastructure for uptime, the design integrates application-level strategies to enhance reliability.
3. Faster Recovery: In case of incidents, the focus shifts to maintaining service continuity rather than restoring the previous state, utilizing Infrastructure as Code (IaC) for quick environment rebuilding.

Monitoring and Automation

To ensure the health of the new cloud, LY Corporation employs various monitoring tools like Prometheus and Grafana. These tools help detect anomalies early, allowing for prompt responses to potential issues. Inoue mentioned that the company automates many processes, from detecting hardware failures to reintegrating replaced components into the clusters.

Security Enhancements

This consolidation comes on the heels of previous infosec problems that exposed user data, prompting government intervention to improve security measures. By streamlining their cloud infrastructure, LY Corporation aims to bolster security and privacy for its users, ensuring compliance with regulatory standards.

What's Next

As LY Corporation moves forward with the Flava cloud, it plans to contribute functional changes back to the upstream OpenStack project. This proactive approach not only enhances their infrastructure but also supports the broader open-source community, fostering collaboration and innovation in cloud technologies.