Heterogeneous Accelerated Compute Cluster
Under the scope of the external pageAMD Xilinx University Program,call_made the external pageHeterogeneous Accelerated Compute Clusters (HACCs)call_made is a special initiative to support novel research in adaptive compute acceleration for high-performance computing (HPC). The scope of the program is broad and encompasses systems, architecture, tools, and applications. HACCs are equipped with the latest AMD Xilinx hardware and software technologies for adaptive compute acceleration research. Each cluster is specially configured to enable some of the world’s foremost academic teams to conduct state-of-the-art HPC research.
Our group hosts one of the five HACCs sites worldwide (the others are two in the US, one in Germany, and one in Singapore).
Infrastructure
ETHZ-HACC comprises high-end servers, Xilinx accelerator cards, and high-speed networking. Each accelerator card has all of its Ethernet interfaces connected to a 100 GbE leaf switch to allow exploration of arbitrary network topologies for distributed computing. Additionally, we are offering a build server with development and bitstream compilation purposes. For more information, please visit the external pageInfrastructurecall_made section on GitHub.
AMD Xilinx hardware and software technologies
ETHZ-HACC offers the following external pageadaptable accelerator cardscall_made:
- external pageAlveo U250call_made
- external pageAlveo U280call_made
- external pageAlveo U50Dcall_made
- external pageAlveo U55Ccall_made
- external pageVersal VCK5000call_made
We use Ansible Automation Platform for managing Xilinx’s tools versioning according to external pageXRT’s release schedulecall_made. All servers equipped with an FPGA are associated with a unique software version, including XRT’s Xilinx Board Utility (xbutil), Xilinx tools (Vivado, Vitis_HLS, Vitis), and the flashable partitions (or base shell) running on the FPGA. For more information, please visit the external pageReleasescall_made section on GitHub.
Hardware acceleration platform
As a multi-core, GPU, and FPGA cluster, ETHZ-HACC allows research in distributed systems, databases, cloud computing, and hardware acceleration of data science. Internally, we are using our ETHZ-HACC to conduct extensive research to assemble an Infrastructure for heterogeneous architectures and hardware acceleration. All our designs are open-source and available through our command-line interpreter external pageCLIcall_made. With this initiative, we hope that ETHZ-HACC users will be able to develop their accelerated applications more quickly.
Get started
Researchers from academy and industry are invited to apply to the HACC program external pageherecall_made. For researchers at ETH Zurich, please contact Prof. Dr. Gustavo Alonso. Once your account has been approved, you will receive a confirmation email with further details.
Tutorials
As part of our partnership program with AMD, we offer a series of tutorials teaching how to use our Hardware Acceleration Platform on the HACC:
People
external pageRuiz, Mario, Dr.call_made
For more information about the people involved in this project, please visit the external pageWho does whatcall_made section on GitHub.