The challenge
On-premises capacity queues delayed research jobs, with no automated scaling or project-level cost attribution.
← All engineering work
Engineering case study
CSIRO
Elastic compute for research workloads.
The engineering
AWS Batch, Spot capacity, Step Functions, S3 and cost attribution by research group.
The result
Self-service job submission with elastic capacity and cost reporting for research teams.
- Client
- CSIRO
- Industry
- Scientific Research
- Location
- Canberra, Australia
About project
Cloud Architecture
AWS Cloud Security
Managed Services
AWS
CSIRO operates large-scale scientific computing workloads across climate modelling, materials science and bioinformatics. tMinus1 designed and delivered the migration of HPC-adjacent workloads to AWS, replacing on-premises batch infrastructure with elastic cloud compute.
The situation
Research teams were queuing for on-premises compute capacity for days during peak periods, delaying experiments and publication timelines. The existing environment had no automated scaling and no cost attribution by research project.
What we built
AWS Batch with EC2 Spot Fleet gave research jobs elastic capacity that scaled from zero in under three minutes. Step Functions orchestrated multi-stage pipeline jobs. S3 replaced NFS for intermediate data, eliminating shared-storage bottlenecks. IAM tagging enforced cost attribution per research group.
How we delivered it
Architecture design and pilot workload (weeks 1–3), Spot interruption testing and retry logic (weeks 4–5), full rollout and researcher onboarding (weeks 6–8).
The result
Queue wait time reduced 60% in the first month. Annual compute cost recovered $1.2M versus on-premises equivalents. Research teams gained self-service job submission with automatic cost reporting to group leads.
More engineering work
Discuss a similar workload.
Request a technical call
Talk with an engineer about your workload and priorities. We explain the scope and cost drivers; a written quote follows once the work is defined.
