Cloud Bursting
Definition
Cloud Bursting is a strategy for temporarily shifting workloads from a private cloud to a public cloud during peak demand, ensuring resource availability.
Use Cases
- Netflix: Handling variable customer demand by running services on elastic public cloud infrastructure rather than fixed on-prem capacity. — Netflix migrated from its own data centers to AWS and uses auto scaling and distributed architectures to add capacity during peak usage periods. (Improved ability to scale for demand spikes and reduced reliance on fixed data center capacity.)
- Adobe: Scaling high-demand digital services by using public cloud elasticity instead of provisioning for peak on-prem usage. — Adobe transitioned major services to a cloud-based model and uses elastic infrastructure to meet variable demand. (Better scalability during peak periods and faster ability to deliver services without overbuilding fixed infrastructure.)
- CERN: Expanding compute capacity for large-scale scientific workloads when local resources are insufficient. — CERN uses a hybrid approach that can extend workloads to external cloud resources during periods of high compute demand. (Access to additional compute capacity without permanently expanding on-prem infrastructure.)
Frequently Asked Questions
- What’s the difference between cloud bursting and auto scaling?
- Auto scaling increases or decreases capacity within the same environment (for example, adding more instances in a public cloud). Cloud bursting specifically means expanding from a private environment into a public cloud when the private environment reaches its limits.
- When should I use cloud bursting?
- Use cloud bursting when you have predictable or occasional spikes (seasonal traffic, batch jobs, end-of-month processing) and you want to keep steady-state workloads on private infrastructure for cost, compliance, or latency reasons—while still having a safe way to handle peak demand in a public cloud.
- How much does cloud bursting cost?
- Costs depend on (1) public cloud compute used during bursts (instances/containers/serverless), (2) storage and database usage if data is replicated or staged in the public cloud, (3) network egress/ingress and private connectivity (VPN or dedicated links), (4) licensing (OS, middleware, commercial apps), and (5) engineering/operations overhead to build and test the hybrid automation. The biggest hidden cost is often data movement and integration complexity, not just compute.
Category: cloud
Difficulty: advanced
Related Terms
See Also