Panels

The first panel is focused on the highlight topic of next year’s cluster conference. The interesting part is that many big data clusters are built from cheap commodity components with many disks per compute node and where the compute nodes are (at best) connected through a 10-Gigabit Ethernet networks. This is not in line with the standard architecture of high-performance clusters, which relies on tight interconnects between the nodes and a parallel file system. The increase of data intensive applications can therefore have a substantial impact on the architecture and the price/performance of next generation cluster computing.