GCVE Advanced Auto-Scaling

Let’s pick up where we left off from last months article and start setting up some of the features of GCVE, starting with Advanced Autoscaling.

What is Advanced Auto-Scaling?

Advanced Autoscaling automatically expands or shrinks a private cloud based on CPU, memory and storage utilisation metrics.
GCVE monitors the cluster based on the metrics defined in the autoscale policy and decides to add or remove nodes automatically. Remember: GCVE is physical Dell Poweredge servers, not a container/VM running in Docker or on a hypervisor like VMware. So provisioning new nodes will take time, though the process is 100% automated.

Note: Autoscaling policies are not available on single node clusters for obvious reasons.

Predefined Policies

GCVE comes with four predefined policies.
Because compute, memory and storage are independant, the policies monitor multiple metrics using OR logic for node addition and AND logic for node removal.

The table below shows the predefined policies:

Autoscale policy Add node condition Remove node condition
Storage Capacity Optimization Policy Storage consumed > 75% Storage consumed < 40%
CPU Performance Optimization Policy CPU utilization > 70% OR Storage consumed > 75% CPU utilization < 45% AND Storage consumed < 40%
Memory Performance Optimization Policy Memory utilization > 80% OR Storage consumed > 75% Memory utilization < 45% AND Storage consumed < 40%
CPU and Memory Performance Optimization Policy CPU utilization > 70% OR Memory utilization > 80% OR Storage consumed > 75% CPU utilization < 45% AND Memory utilization < 45% AND Storage consumed < 40%

Updating an autoscale policy

To view or change an autoscaling policy associated with a cluster, follow the below steps:

  • Access the VMware Engine portal
  • Goto Resources on the left menu and select a private cloud
  • Goto Clusters and click the Edit icon at the right of the cluster line
  • On this page, you can view the current policy applied to your cluster, select a new one from the drop-down list or optionally adjust policy details
  • Click Submit to complete the adjustment.

Definitions that will be helpful:

  • Add nodes: Condition that must be met before GCVE will add a node to the cluster.
  • Remove nodes: Condition that must be met before GCVE will remote a node from the cluster.
  • Cool off period: Minimum amount of time between actions taken due to autoscaling policy.
  • Node limits: Minimum and maximum number of nodes in a cluster for autoscaling.

Summary

Autoscaling policies are a powerful tool to ensure you are paying the least amount of money per month for the highest availability within your private cloud whilst ensuring the cluster maintains its SLA. Used correctly, you will always have the minimum amount of nodes whilst keeping your uptime at or above 99.99%.