Managed Service for Apache Cassandra

  • Score Awaiting client review
  • Date Published
  • Reading Time 2-Minute Read
Data engineering case study banner

Managed Service for Apache Cassandra.

Client: Large telco company

Project Duration: 6 months (ongoing)

Goal: Create fully managed multi-region Apache Cassandra cluster deployed in cloud (AWS) for client so their development team can focus on business functionalities

Tech: Apache Cassandra, AWS, DataDog, ELK, Ansible

The Challenge

Client developed a data model for multi-region initial use case on top of Apache Cassandra cluster but soon they realized that operating, monitoring and scaling large multi-region cluster in the cloud takes tame and skills. We jumped in, defined SLA and took over operation of the cluster so client’s technical team can focus of further development of business functionality.

The Approach

When we took over, Apache Cassandra was installed manually on each instance in multiple AWS regions, and monitoring was pretty basic. We suggested first to do production readiness engagement (assess current configuration, do load testing, failure scenarios and automate ops) before moving comfortable to managed service engagement according to defined SLA. Managed service mode included maintenance and support mode, overlooking whole cluster, expanding if needed into new regions (or expanding size in current regions) and creating data models for new functionality.

The Solution

Our solution is fully managed cluster with entry point for development team. We automated monitoring and alerting for our ease of operations but we also provided access for client if someone from the team wants to follow performance. We are practicing constant failure testing and stress testing so we can be confident that cluster is performing according to SLA.

The Results

Result is fully managed Apache Cassandra cluster spread into multiple regions, with 24/7 support according to defined SLA. Result is also limited pool of hours of our service to create new data models or fine-tune cluster performance. Last but not the least most important result is rapid development of application itself since clients technical team can now focus on business features.