Sviatoslav Stumpf and Vladislav Povyshev, ITMO University, RUSSIA

The aim of this paper is to examine and demonstrate how integer-based datetime labels (integer surrogate keys for time) can optimize data-warehouse and time-series performance, proposing practical formats and algorithms and validating their efficiency on real-world workloads. It is shown that replacing standard DATE and TIMESTAMP types with 32- and 64-bit integer formats reduces storage requirements by 30–60% and speeds up query execution by 25–40%. The paper presents indexing, aggregation, compression, and batching algorithms demonstrating up to an eightfold increase in throughput. Practical examples from finance, telecommunications, IoT, and scientific research confirm the efficiency and versatility of the proposed approach.

integer labels, time series, optimization, performance, data warehouse, indexing, aggregation