Glossary · Critical cloud · Elastic compute
Cloud bursting: absorbing compute spikes without over-provisioning
Cloud bursting means keeping the base compute load on your own infrastructure and offloading only the spikes to the public cloud: the month-end risk close, a Monte Carlo run, a one-off actuarial calculation. It avoids over-provisioning the on-premise grid for demand that occurs only occasionally, without giving up capacity when you need it.
When it fits
- Stable base load — recurring, predictable compute stays on the owned grid, which is already amortized.
- Occasional spikes — risk closes, regulatory scenarios or actuarial batches that drive demand up for a few hours or days.
- Parallelizable workloads — Monte Carlo, portfolio valuation, capital calculation: tasks that spread across many nodes and scale well.
- Non-critical data in transit — or properly protected and anonymized before it leaves the perimeter.
Why it matters in banking and insurance
In HPC workloads (Monte Carlo, FRTB, XVA, actuarial calculation), sizing the on-premise grid for the peak means paying all year for capacity that is used only a few days. Cloud bursting shifts that peak to the cloud and is governed as a FinOps decision: cost per unit of compute and per scenario, not by default. The regulatory flip side is DORA: any offloading to a cloud provider falls within the operational resilience perimeter and requires reversibility and a documented exit strategy.
How Vermont Solutions helps
Hybrid HPC grid with governed bursting
Vermont designs grid architectures that combine on-premise capacity with elastic bursting to the cloud, sized by real cost per scenario and with the reversibility that the financial regulator requires.
See HPC and Grid Computing →Fuentes
Last updated: 2026-06-21. Editorial content by Vermont Solutions, citable with attribution.