Abstract:
The scale of cloud storage systems is extensive and rapidly evolving. Nowadays, a tiered cloud storage system is in high demand amongst storage providers. Tiering is a well-known solution which is used to optimise the usage of data storage disks to gain better than average access speed for any storage system. A framework for benchmarking storage tiers is required to investigate the utilisation of storage infrastructures. In the research discussed in this dissertation we developed a benchmarking framework for testing tiered cloud storage. The benchmark utilises non-archival storage tier workloads modelled from daily traces of computing systems and typical file size distributions. Considering the generation of a realistic workload, we scaled our workload model to match contemporary data storage usage. In order to do the benchmarking we extended an existing benchmarking tool to test our tiered, cloud storage setup with the modelled workload. We use these benchmarks to detect contention for storage system infrastructure when there is no migration between tiers. Testing the tiers demonstrated the contention between our resources because the system did not perform as well as expected.