ScyllaDB on Kubernetes: How to Run Intense Workloads with Spot Instances

Visually.io has transitioned from MongoDB to ScyllaDB for its main production real-time database, leveraging Kubernetes and spot instances to manage intense workloads at a quarter of the usual cloud compute cost. ScyllaDB, an open-source NoSQL database compatible with Apache Cassandra and DynamoDB, offers advantages such as zero downtime failovers, horizontal and vertical scaling, and efficient data sharding, which ensure high availability and low latency even when subjected to frequent failovers inherent to spot instances. The platform's architecture avoids common issues associated with Cassandra and enables seamless integration into Kubernetes environments through its Helm chart, operator, and manager, which automate deployment and management tasks. Monitoring is facilitated by Prometheus and Grafana, providing insights into metrics and performance, while Kubernetes ConfigMaps enable the creation of dashboards for detailed database monitoring. Despite the occasional latency spikes due to the volatile nature of spot instances, the cost savings and operational efficiency make ScyllaDB a compelling choice for scalable database management.