Dolphin Interconnect Solutions’ eX3 products are used to connect computers and IO systems together to create high performance computing platforms. These clusters have faster response times and transaction throughput than systems built using more conventional technologies.
Dolphin provides the SIRIUS laboratory with a composable infrastructure with high performance and low latency. Graphical processing units (GPUs), storage units using the Non-volatile Memory Express (NVMe) standard and CPU-based systems can be connected together and dynamically configured for high application performance. Dolphin’s use of the Peripheral Component Interconnect Express (PCIe) standard allows resources to communicate at native speed without software overhead. The platform also supports traditional clustering software like MPI and Berkeley socket communications.
Our work in SIRIUS is to continue to develop an infrastructure that supports pools of hundreds of compute resources and CPU systems in a virtualized environment. Fast access to storage will drastically reduce the time needed to load and store data from NVMe devices. Optimized integration of GPUs and HPC libraries will ensure the best possible performance for resource-demanding Big-Data applications.
The eX3 infrastructure will be an important prototyping environment for large scale testing and benchmarking of Dolphin’s infrastructure with real world applications from SIRIUS’ partners.