3.2.2 Scaling Management
When scaling out, the orchestration of the scaling process needs to not only configure the resources, but to also manage the scaling task. This includes activities such as:
Verifying available capacity prior to consuming additional resources.
Coordinating or restricting parallel scaling activities on the same application.
Generally, scaling activities should be serialized to avoid issues that can arise from the parallel execution. For example, in a parallel operation you must manage name and IP assignment across tasks to prevent duplicate names or IP addresses from being assigned to the new resources. Additionally, parallel deployment across tiers can result in over-shooting the KPI goal, resulting in the immediate triggering of a scale back activity. Parallel scaling can be implemented, but due to the complexity, it should be implemented after serial scaling has proven stable.
Figure 12. Scaling Management
The scaling management process is driven by directives contained within the application metadata itself or by information contained within the scaling workflow. In this example several directives are contained as XML data within the OVF descriptor.
<ns2:Property ns2:value="4" ns2:userConfigurable="true" ns2:type="string" ns2:key="autoscaleMinTier1Instances">
<ns2:Label/>
<ns2:Description>Minimum number of instances for tier 1 VMs</ns2:Description>
</ns2:Property>
