PGDay Lowlands2024

Time series, essential in diverse fields from IoT to finance, pose unique storage and querying challenges. At OVHcloud, a major cloud provider in Europe, we've leveraged the TimescaleDB extension to manage this type of data more effectively.

This talk will begin by outlining the basics of time series and the key features of TimescaleDB.

We will then discuss the use cases and challenges we've encountered including:

• Performance issues: which parameters to tune for write-heavy workloads.
• Continuous aggregates: effectively create bills with billions of data points and predict the future.
• Compression strategy: improving disk space and response times by choosing the appropriate segment columns.
• Schema migrations: how to add columns without blocking the service.
• Change Data Capture: real-time streaming of data points into the Data Lake infrastructure.
• Observability benchmark: send Prometheus metrics of thousand of servers.
• Zabbix tuning: how to effectively remove old data points from existing deployments.

In this user feedback session, we'll share our successes and failures in using TimescaleDB at scale, with practical examples from real-world deployments.