UPDATE 07/14/25—Added Q&A Set #1
This website enables public access to Request for Information No. HPC-007 (RFI) pertaining to a Future Generation High Performance Computing Center. The RFI points of contact are LLNS Contract Analyst Gary Ward (ward31@llnl.gov) and Distinguished Member of Technical Staff Dr. Todd Gamblin (gamblin2@llnl.gov).
The RFI is released on 06-25-25. The RFI consists of documents available at this website. RFI Respondents may review the documents and submit corresponding responses. LLNS does not anticipate any future updates to the RFI; however, interested RFI Respondents are advised to monitor this website (at their discretion) for any potential future updates or other related RFI information.
RFI Documents:
- RFI Letter
- Attachment (1) RFI Technical Considerations
For background information about high performance computing at the Lawrence Livermore National Laboratory, please visit the links below for related information:
- https://computing.llnl.gov/livermore-computing
- https://hpc.llnl.gov/hardware/compute-platforms
- https://hpc.llnl.gov/sites/default/files/LC-facilities-resources.pdf (printable)
Question & Answers Set #1
QUESTION 1
Is there an expectation to include Key Personnel in RFI responses? If not, can LLNL confirm that whether Key Personnel resumes are still required in RFI responses?
ANSWER 1
There is no requirement to list key personnel — this is an RFI aimed at information gathering. The goal of the RFI is for responders to describe approaches that LLNL could consider for future procurements and why LLNL should consider them.
QUESTION 2
Can LLNL confirm what percentage of current workloads require?
- Classified processing
- Real-time/near-real-time processing
- Specific hardware acceleration
- Bare metal performance
- Multi-site collaboration
ANSWER 2
A majority of LLNL current workloads require classified processing. Few of our workloads require real-time processing. LLNL does not require specific hardware acceleration but we leverage hardware acceleration extensively throughout our HPC center, particularly through AMD and NVIDIA GPUs. LLNL requires that virtualized workloads achieve performance approaching that of bare metal although LLNL does not require bare metal specifically. Some workloads require multi-site collaboration although not a majority. LLNL expects the percentage that require it to increase in the future.
QUESTION 3
Can RFI respondents propose off prem or partial off prem solutions?
ANSWER 3
Note that the RFI is not an RFP (i.e., not a Request For Proposal). With that said, yes, RFI respondents may describe off prem or partial off prem solution approaches.
QUESTION 4
LLNL RFI states system procurements will be multivendor "to include HPC/Cloud hardware from a wide range of vendors". Can LLNL please further clarify this multivendor vision?
ANSWER 4
Across the breadth of Livermore Computing (LC) procurements, LLNL requires the ability to use different vendors. The future procurement strategy anticipates individual procurements that each are best addressed through multiple awards to distinct vendors. The goal of this RFI is to understand what technologies will be available from vendors in the 2030 timeframe and why LLNL should consider them for its systems. LLNL desires to select future procurement elements that, together, maximize the value of the HPC center, and this RFI describes a procurement model that enables selecting optimal elements even if offered by different vendors. LLNL is interested in how vendors see themselves fitting in to this model.
QUESTION 5
Would LLNL consider hardware-based systems designed for on-premises deployment but originating from commercial cloud environments that provide cloud-native management interfaces (e.g., APIs for compute, storage, and network orchestration)?
ANSWER 5
Yes, LLNL would consider hardware-based systems designed for on-premise deployment but that use interfaces originally designed for commercial clouds.
QUESTION 6
Would LLNL consider hardware-enforced trusted execution environments (e.g. secure enclaves) as acceptable methods to support the FG-HPCC's need for strong on-node isolation and multi-tenant support; particularly for untrusted CI workflows or persistent services?
ANSWER 6
LLNL is very interested in hearing about secure enclaves and methods to support isolation.
QUESTION 7
Is LLNL open to lightweight virtualization technologies (e.g., microVMs)?
ANSWER 7
Yes, LLNL is open to lightweight virtualization technologies. A goal this RFI is to understand the range of such technologies that will be available and how they would support future HPC systems.
QUESTION 8
Can LLNL provide specific latency and bandwidth requirements for?
- Node-to-node communication
- Storage access
- Cross-site data movement
ANSWER 8
A goal of this RFI is to understand the technologies that will be available in the 2030 timeframe and the latencies and bandwidths that they will support.
-end-