|Name:||(03) Compression By Default – Reducing Total Cost of Ownership of Storage Systems|
|Time:||Monday, June 23, 2014
04:36 pm - 04:43 pm
CCL - Congress Center Leipzig
|Speaker:||Michael Kuhn, University of Hamburg|
|Abstract:||Both energy and storage are becoming key issues in high-performance (HPC) systems, especially when thinking about upcoming Exascale systems. The amount of energy consumption and storage capacity needed to solve future problems are growing rapidly; the HPC community must face this problem in cost-/energy-efficient ways. In this poster, we provide a model to estimate the TCO of HPC storage systems. We use the Lustre parallel distributed file system and enable additional data processing on the storage servers to compress all file system data. Lustre's ZFS backend, which natively supports compression, is used to show that data compression can help alleviate capacity and energy problems. Previous results demonstrate that compression ratios of more than 1.5 are possible for scientific climate data with only negligible increases in CPU utilization and power consumption. Our model shows that these low-overhead algorithms can help reducing the TCO of storage systems by roughly the same factor.
Michael Kuhn, Konstantinos Chasapis, Manuel Dolz & Thomas Ludwig, University of Hamburg
Michael Kuhn: firstname.lastname@example.org