Performance
We regularly test our architecture, infrastructure, and query engine at scale to ensure Altertable delivers consistent performance as your data grows.
Benchmarking
One test we've heavily relied on so far has been TPCH with a scale factor of 100. This is a 250GB dataset that provides a comprehensive workload for evaluating query performance. On this benchmark, Altertable does very well at a fraction of the cost compared to traditional analytical databases.
Query Time
Average TPCH Benchmark (sf=100) results - lower is better
Cold vs Hot Queries
Due to our lakehouse architecture, some data might be required to get downloaded from the infinite-scale distributed storage to the local DuckDB worker machine. Therefore, the first query might be considered "cold"—most of the elapsed time will probably be spent downloading the data. However, subsequent queries (hot) will be significantly faster as the data is already available locally.
Our Approach to Performance
It's important to note that each benchmark can always be tricked and tuned to say whatever we want. This is not something we're optimizing for right now. Instead, we're focused on testing with our customers' real-world scaling problems and ensuring Altertable performs well under actual production workloads.