4.3 Cloud BurstFrom Data Center to Cloud
Actors: cloud-subscriber, cloud-provider, cloud-management-broker
Goals: Maintain required service levels for an agency's data-center hosted process, by dynamically allocating/deallocating cloud computer or storage resources to service current demands.
Assumptions: Assumes the Use case "Open an Account"
Success Scenario 1 (base, IaaS):
Cloud-subscriber provisions and maintains cloud-provider-1 virtual machine images and/or configured storage capacity designed to support cloud-subscriber-defined units of work ranging in scope from individual computing or storage tasks to entire distributed applications.
Cloud-subscriber establishes load monitoring processes for the units of work concerned, and load threshold and sensitivity limits for cloud bursting. Upper limits govern starting new processes on cloud-provider to handle increasing load; lower limits govern stopping cloud-provider processes to handle decreasing load.
As monitored load triggers threshold limits, processes start or stop on cloud-provider-1 infrastructure to maintain required service levels.
Failure Conditions 1 (base): Failed allocation or deallocation event
Failure Handling 1 (base): Failed allocation or deallocation event: Cloud-provider-1 notifies cloud-subscriber. Cloud-subscriber either communicates with cloud-provider-1 for resolution within an acceptable SLA or has access to automated cloud-provider-1 notification and resolution. Failed deallocation events can result in excess agency charges and must be covered in SLA agreements.
Success Scenario 2 (Manual Bursting, IaaS): Cloud-subscriber manually allocates and deallocates cloud-provider resources based on threshold notifications.
Failure Conditions 2 (Manual Bursting): N/A
Failure Handling 2 (Manual Bursting): N/A
Success Scenario 3 (Automated Bursting, IaaS): Cloud-management-broker processes monitor load and threshold limits and allocate or deallocate cloud-provider-1 resources using programming interfaces provided by cloud-provider.
Failure Conditions 3 (Automated Bursting): Failed event detection
Failure Handling 3 (Automated Bursting): Cloud-management-broker independently monitors its event detection services and notifies cloud-subscriber of outages so cloud-subscriber can fall back to manual bursting scenarios.