X-Mem: A Cross-platform and Extensible Memory Characterization Tool For the Cloud

Effective use of the memory hierarchy is crucial to cloud computing. Platform memory subsystems must be carefully provisioned and configured to minimize overall cost and energy for cloud providers. For cloud subscribers, the diversity of available platforms complicates comparisons and the optimization of performance. To address these needs, we present X-Mem, a new open-source software tool that characterizes the memory hierarchy for cloud computing. X-Mem is designed to be modular, portable, and extensible while surpassing most capabilities of existing utilities. The tool directly measures a number of statistics for throughput, (un)loaded latency, and power for each level of cache and DRAM through flexible stimuli. Its features include multithreading, awareness of non-uniform memory architecture, and support for different page sizes. X-Mem can exercise memory using many combinations of load/store width, access pattern, and working set size per thread. The accessibility and extensibility of our tool also facilitates other research purposes. We demonstrate the utility of X-Mem through a series of experimental case studies using state-of-the-art platforms. Our results show how cloud subscribers could choose a preferred target platform and better optimize their applications even if the hardware/software stack is opaque. Cloud providers could use X-Mem to fine-tune system configurations and to verify machine performance envelopes before deployment. We envision novel ways that researchers could extend X-Mem for purposes such as the characterization of emerging memory architectures.