Towards an Efficient Benchmark Generation Engine for Garbage Collection

ACM/SPEC International Conference on Performance Engineering Companion, Berlin, Germany, April, 2018Garbage Collection (GC) is a key feature of many modern programming technologies. It offers significant software engineering benefits over explicitly memory management. Nonetheless, it is also a major cause of performance degradation. As the rate of adoption of GC-related technologies continues to grow, it is highly relevant to understand its performance impact. However, this is challenging due to the non-deterministic nature of GC. To tackle this problem, we present an engine (HERMES) to create realistic GC benchmarks by effectively capturing the GC/memory behaviours exhibited by real-world Java applications. Our experiments prove how HERMES can be useful to strengthen the evaluation of GC-related advancements. This is achieved by broadening the number and diversity of the test scenarios, as well as reducing the time invested in testing.Science Foundation Irelan