Rightsizing Configuration Defaults
Exostellar’s Workload Optimizer (WO) automatically scales your workloads based on CPU and memory utilization, using thresholds and cooldown periods to ensure stable and responsive performance. Below are the default behaviors built into the system:
Scaling Thresholds
WO monitors utilization metrics and makes scaling decisions based on sustained resource usage. The following default thresholds apply to CPU and memory:
Event | Threshold |
|---|---|
CPU - Scale Down | 60% |
CPU - Scale Up | 80% |
Mem - Scale Down | 60% |
Mem - Scale Up | 80% |
Cooldown Periods
To avoid rapid oscillation between scaling actions, WO enforces cooldown periods after each adjustment:
Event | Cooldown Duration |
|---|---|
After Scaling Up | 10 minutes |
After Emergency Scale-Up | 15 minutes |
Once a scale-up occurs, WO waits 10 minutes before evaluating whether to scale down again. For emergency scenarios (described below), a longer 15-minute cooldown ensures workload stability.
Explanation:
If utilization drops below 60%, WO will initiate a scale-down action and ensure a 10-minute cooldown before considering any scale-downs.
If utilization remains above 80%, WO will initiate a scale-up action and ensure a 10-minute cooldown before considering any scale-ups.
These thresholds allow workloads to dynamically adapt to changing demand.
Emergency Scale-Up
In scenarios of sudden, high utilization, WO includes an emergency scaling mechanism to preserve workload performance:
Condition | Action |
|---|---|
Utilization is below 50% | WO scales up to 50% of the Max Mem immediately |
Utilization exceeds 50% | WO scales to the next instance size |
This emergency response ensures that workloads experiencing rapid spikes in demand are handled without delay.