Application Bursting
Overview
This cloud approach replicates a full application capability in the cloud, but the cloud capability is only used for burst events where the application temporarily needs additional capacity - perhaps at month-end or for bursts of end-user customer demand.
Use
In practice, this approach is much harder than it first appears, and is virtually impossible for any application that has not been specifically built to support bursting. Whilst the application middle-tier is easy to scale (if it is essentially stateless, that is) the data tier provides is a much greater challenge; because of the necessary data integrity and consistency requirements, it is usually infeasible to split an application datastore between two databases, and even if is possible (e.g. with sharing), then regressing back to a single database when the burst capacity is no longer needed is equally challenging.
Some applications, however, do lend themselves to this type of bursting - typically where there is no need for complex business transactions, or no data that requires transactional integrity.
In general, if the app was built to enable this from the design then it’s possible; don’t try this for legacy monolithic apps - try component bursting instead.
Use Cases :
- HPC
- All types of computation where there is an input dataset and an output dataset and multiplexed calculations can be done in isolation
- Algorithmic Calculation
- Financial risk calculations, scientific modeling, engineering simulation
- Read-only services
- Product Catalogues, Media services (inc streaming)
- Isolated business process or sub process
- Application/quoting processes (e.g. mortgage/insurance)
- Fully devolved (to user level) services
- E.g. personal file storage, ToDo managers,