With business growth so tightly coupled to the performance of its database infrastructure, the organization chose Cassandra to help them meet their mandate of developing usage profiles and growing ad revenue. Using in-house resources to design and implement the system, they ran into a number of issues and realized they lacked the internal knowledge to resolve them. Working with Pythian’s Cassandra team they were able to develop a scalable, resilient, distributed Cassandra database infrastructure that quickly and efficiently processes data requests. One of the world’s leading business news organizations set a new standard for news delivery when they implemented a multi-tiered subscription program for web-based access to its news material. With 4.5 million registered users and over 600,000 paying subscribers, the organization relies on its database infrastructure to develop in-depth usage profiles of its membership to understand key interests and to grow advertising revenue. With business growth so tightly coupled to the performance of its database infrastructure, the organization chose Cassandra because of its strength in supporting geographically disperse infrastructures and for its ability to scale. They were using in-house resources to design and implement the system, but soon ran into a number of performance issues and realized they lacked the internal knowledge to fully troubleshoot and resolve them. Cassandra often requires manual effort to ensure data is synchronized across all servers. In addition to performing manual repair operations on a weekly basis, the organization was also having trouble bringing new servers online because data was not being totally transferred to the new servers – in some cases less than 10%. Compounding the problem was data models that did not properly distribute query requests across all the servers. Instead of the query being routed using the shortest path to the proper server, the query would be sent on a roundabout path that touched almost every server in the network before reaching its intended destination. Performance was severely degraded, and the infrastructure was not able to scale.

The organization turned to Pythian who could draw upon broad experiences with other Cassandra clients to help troubleshoot and resolve the root cause of the recurring problems. The first priority was to determine why the new servers were not coming online correctly, and why the replication mechanism wasn’t working properly. Cassandra is very sensitive to cloud-based hardware configurations, and spinning up new servers and manual repair processes are extremely taxing on the hardware and network links. After a comprehensive “deep dive” analysis, Pythian determined that certain network conditions caused interruptions within the hardware stack. Pythian recommended a software upgrade and some changes to the configuration to improve network throttling. Once performance issues were stabilized, Pythian then examined the data modeling architecture and proposed a series of recommendations to fine-tune and optimize system availability and performance.