Benchmarks for hybrid transactional/analytical processing (HTAP) are designed to evaluate systems that can efficiently handle both real-time transactions and analytical queries simultaneously. Instead of separating these two workloads, HTAP benchmarks create scenarios where transactional data is processed instantly while also allowing for complex queries and data analytics on the same dataset. This approach provides a more accurate representation of how systems behave under the mixed load typical of modern applications.
One widely recognized benchmark for HTAP is the TPC-E benchmark, which simulates a real-time trading environment where transactions are processed continuously while also allowing for in-depth analysis of trading activities. This benchmark emphasizes the need for low-latency responses on transactional workloads while ensuring that analytical jobs can run without degrading the performance of real-time operations. By incorporating both types of workloads, it helps developers gauge the performance of their systems under realistic conditions.
Another important aspect of HTAP benchmarks is their focus on operational simplicity and resource efficiency. For instance, benchmarks like the YCSB (Yahoo! Cloud Serving Benchmark) focus on key-value stores that might combine transactional and analytical tasks in a cloud environment. They evaluate how well a system can provide quick read and write operations while also handling larger analytical queries. In these scenarios, it's crucial to measure not just transaction times, but also how quickly a system can return results from complex queries, thus offering a comprehensive view of its performance capabilities in HTAP contexts.