Oversubscription metrics

ThinkAgile CP offers oversubscription at two levels.

• Migration zone – Migration zone oversubscription is set as virtual data centers are created.

• Node – Node-level oversubscription is set as application instances are provisioned inside virtual data centers.

Both levels are largely static quantities whose values are a function of the virtual data centers created and the applications provisioned. Additionally, ThinkAgile CP also maintains the runtime load at the migration zone and node levels. These levels are dynamic quantities whose values change with time and depend on what resources the applications actually consume when running. The migration zone level runtime load is a weighted average of the various node level runtime loads. ThinkAgile CP maintains both compute and memory loads at the migration zone and node levels.

Note

When a VM, or application instance, is first started in ThinkAgile CP, it consumes 100% of its provisioned memory resources. Then, over time (usually within a few minutes), the memory usage will scale back to what the instance is using. Because of this, an instance can only be started on a compute node if that node has enough available memory for the instance's provisioned memory.

• Oversubscription at migration zone level
  Oversubscription at virtual data center creation time is calculated by category across a migration zone. Virtual data centers are allocated inside migration zones, and oversubscription is calculated after the virtual data centers are created.
• Oversubscription at the node level
  Next, we provision application instances inside virtual data centers. An application instance must specify the virtual datacenter it runs in. It also specifies the amount of resources it needs, and whether those resources must come from a certain category or from any category.