Higher Education & Research
Peregrine complements existing university HPC, institutional GPU clusters and cloud research environments with Australian GPU infrastructure — owned by the institution or supplied as capacity — and operates the platform layer researchers depend on, with the Slurm and Kubernetes interfaces they already know.
Infrastructure can be designed to support customer security, governance and compliance requirements. Use one service or all four.
Part 1 · Use cases
Use case 01
Training and adapting large models for research requires tightly coupled multi-GPU nodes and long, uninterrupted jobs.
Reserved nodes with fair-share scheduling give research groups predictable access without institutional queue times.
GPU considerations
Use case 02
GPU-accelerated numerical, climate, physics and engineering simulation workloads sit alongside AI on the same infrastructure.
Partitions and quotas let HPC and AI users share GPUs fairly.
GPU considerations
Use case 03
Sequence analysis, microscopy, remote sensing and medical imaging research push large datasets through GPU pipelines.
Storage throughput is sized together with GPU count in the Compute Blueprint.
GPU considerations
Use case 04
Policy learning, simulation and generative model research are bursty around publication cycles.
Project capacity on Peregrine infrastructure complements the owned cluster when deadlines approach.
GPU considerations
Part 2 · How to deploy with Peregrine
Size the research workload mix and compare an owned cluster, Peregrine capacity and cloud research credits.
Learn moreDesign, source and commission a university-owned GPU cluster — Slurm or Kubernetes, shared storage, quotas — in your facility or an Australian data centre.
Learn moreOperate the platform: multi-tenancy, grant and project-based allocation, quotas, user management and monitoring.
Learn moreBurst and project capacity on Peregrine infrastructure when the owned cluster is full or a grant needs GPUs quickly, subject to availability.
Learn moreRecommended platforms
Indicative only; the Compute Blueprint confirms platform, interconnect and scale per project.
Hopper
Proven production platform with the broadest software maturity.
View platform
Hopper
Mature ecosystem with larger memory for inference of large models and training.
View platform
Blackwell
Balanced Blackwell platform for training and production inference.
View platformInfrastructure models: capital vs operating
Universities typically fund owned clusters through capital and infrastructure grants, while project capacity aligns with research grants and operating budgets. Peregrine can structure an owned build with Peregrine operating it, supplemented by reserved or project capacity — scoped per engagement, with no prices published here.
FAQ
Tell us about your higher education & research workloads and deployment preferences. A compute specialist will discuss the appropriate infrastructure model.
