Motivation

SUTS provides the government with a tool to develop transportation polices and plan for future infrastructure. It helps citizens organize their daily intra-city trips.

We select five representative programs from SUTS to build a Big Data benchmark suite. We hope it could be used as valid tool to benchmark and optimize such kind of systems.

Key Features

SZTS differs from other benchmark suites in the following ways:

(1)Application Domain

SZTS is derived from the Smart Urban Transportation System of Shenzhen. Its application domain is city transportation systems

(2)Characterization method

We characterize SZTS benchmarks by a cross-layer methodology, we select metrics from job layer and micro-architecture level. Other BigData benchmark suites perform characterization in job level or micro-architecture level respectively.

(3)Input data

We adopt real-world GPS records and smart card transaction data, not synthetic data.

(4)How to select workload

We adopt statistics techniques such as Principal Component Analysis (PCA) and Clustering to select representative workloads and associate input data sets. Unlike other Big Data benchmark suites, Manually Select workload according to the experience of the designers.

Acknowledgements

We would like to acknowledge the hard work of the authors of the benchmark programs.

The contributors of SZTS are different research groups in Cloud Computing Center, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences