Compute for demanding workloads
HPC clusters for research and advanced computing, including specialized environments. PGTEK is one of the few delivery partners with the engineering depth to do this work.
Compute at the edge of the mission
HPC Cluster Design & Build
Full-stack cluster design: interconnect topology, storage architecture, job scheduler configuration, and commissioning.
GPU-Accelerated Computing
NVIDIA DGX and HGX systems for simulation, modeling, and ML workloads. Dense compute with the networking to match.
High-Speed Interconnect
InfiniBand and high-bandwidth Ethernet fabric design and deployment. Interconnect bottlenecks end jobs before they finish.
Parallel Storage Architecture
Lustre, GPFS, and NVMe-oF storage architectures sized for parallel I/O workloads. Storage that keeps up with the cluster.
Classified HPC Environments
HPC delivery inside classified and sensitive compartmented environments. PGTEK holds the clearances to support national security computing programs.
Workload Management & Optimization
Slurm and PBS Professional configuration, queue policy, and tuning to keep utilization high and jobs moving.
Few can do this work
Classified HPC demands clearances, fabric expertise, and a record of delivery. We hold all three.
Top-5 pedigree
In partnership with HPE, PGTEK contributed to the build of 4 of the top 5 HPC supercomputers ever benchmarked.
Cleared to build
Engineers with active clearances for sensitive compartmented and classified programs.
Fabric-first
Interconnect and parallel storage designed so the cluster never waits on itself.
Tuned for throughput
Scheduler and queue policy that keeps utilization high and science moving.