New York State Contract Reporter #27-04

Roswell Park Cancer Institute seeks high performance computing solution for research environment

Roswell Park Cancer Institute Corp. is seeking proposals for a high performance computing (HPC) solution to support its shared research environment for academic groups in translational science, bioinformatics, biostatistics, genomics, and machine learning. - Buyer: Roswell Park Cancer Institute Corp., Purchasing division - OEMs highlighted: NVIDIA (for GPU hardware) - Products requested: - HPC solution with baseline of 21 CPU nodes and 5 GPU-enabled nodes - GPU nodes must include NVIDIA A100 and V100 GPUs - Storage solution approaching 1 petabyte of high throughput (non-cold) capacity - Requirements: - Solution must handle approximately 100,000 CPU core hours annually - Support both CPU and GPU workloads - Accommodate Slurm-scheduled jobs, R, Python, and genomics pipelines - Allow for future growth in GPU demand, storage, and workload diversification - Security requirements may include handling PHI and sensitive data in the future - Respondents must map their configurations to baseline specs and explain differences in concurrency, memory per core, queueing, GPU availability, and storage performance - Preference for predictable, budgeted cost models - Brokers or consultants are not preferred unless required for solution delivery

Description

Roswell Park Cancer Institute Corp. is seeking a partner to provide a high performance computing solution to support multiple academic research groups in translational science, bioinformatics, biostatistics, genomics, and machine learning. The solution should accommodate approximately 100,000 CPU core hours annually, with workloads primarily CPU-based and growing GPU-accelerated machine learning and AI workflows. The environment must support queueing with reasonable wait times and provide storage approaching one petabyte of high throughput. The institution values predictable, budgeted cost models and seeks a partner capable of recommending right-sized architectures with flexibility for future growth.

View original listing