Evaluating models of memory allocation

Abstract

Because dynamic memory management is an important part of a large class of computer programs, high-performance algorithms for dynamic memory management have been and will continue to be of considerable interest. The goal of this research is to explore the size and accuracy of synthetic models of program allocation behavior. These models, if accurate enough, proved an attractive alternative to algorithm evaluation based on trace-driven simulation using actual traces. Based on our analysis, we conclude that even relatively simple synthetic models can effectively emulate the allocation behavior of well-behaved programs. However, even the most complex models we investigate can only roughly approximate the behavior of more complex programs and/or allocation policies. While synthetic models have been used to evaluate the performance of dynamic memory management algorithms, our results show that these models can be inaccurate and must be used with care. Given current trends toward more complex applications and allocation algorithms, the synthetic models we investigate are likely to be even less accurate in the future.