Description

Provides overview of the concept of horizontal auto scaling

Content / Solution:

Horizontal Auto Scaling

Horizontal Auto Scaling allows a user to create a set of rules to start or stop a previously-provisioned Cloud Server assigned to a VIP when a pre-defined monitoring threshold is breached. This allows users to scale the number of servers servicing a VIP up or down based on monitoring results. Horizontal Auto Scaling will not dynamically provision new Cloud Servers. It simply starts and brings into service previously-deployed Cloud Servers that are sitting in a Stopped state (and vice versa).

How To Use Horizontal Auto Scaling

To take advantage of Horizontal Auto Scaling, you will need to deploy two or more Cloud Servers in your Cloud environment.  Although the servers can be running or stopped, a stopped server cannot generate the utilization necessary to trigger a Horizontal Auto Scaling rule.  Since Horizontal Auto Scaling is based on the aggregate utilization of all of the running Cloud Servers in the server pool, you will want to ensure that at least one server is running when you create the rule.

Once the servers have been deployed, you will need to enable Cloud Monitoring for each server.  See the following article for details on how to enable Cloud Monitoring:

You will also need to set up load balancing for your deployed servers.  Otherwise, the servers will not be visible to the Auto Scaling rule.  See the following articles:

Once you have deployed your Cloud Servers and assigned them to the same VIP or Virtual Listener, you will need to create a new sub-administrator account with the 'Network' and 'Server' roles in the Admin UI.  This account will be used to call the Cloud APIs that perform the Horizontal Auto Scaling actions.  We strongly recommend you create a new sub-administrator specific to your auto-scaling effort. Since Auto Scaling leverages our existing Cloud APIs, you therefore will be able to track all of the Auto Scaling activity (e.g. start server, stop server, etc.) tied to this account in the Administrative Logs found in the Admin UI.

You will be prompted to add the username and password to the Auto Scaling Manager before you can create any Auto Scaling rules.  (If you have already added sub-administrator credentials to the Auto Scaling Manager, then you can bypass this step.)  See the following article for details on how to create a sub-administrator account:

You are now prepared to create your Horizontal Auto Scaling rule in the Cloud Monitoring portal.  See the following article for details on how to create a Horizontal Auto Scaling rule:

When a Horizontal Auto Scaling event is triggered based on the thresholds specified in the rule, the monitoring system identifies all running and stopped Cloud Servers that have been assigned to the VIP.  If the Auto Scaling rule was triggered by a breach of the maximum Auto Scaling threshold, then the system will start and bring into service any one of the Cloud Servers sitting in a Stopped state.  Conversely, if the Auto Scaling rule was triggered by a breach of the minimum Auto Scaling threshold, then the system will bring out of service and stop any one of the running Cloud Servers.  It is important to note that these servers are selected from within the server farm by the system at random.

The system will continue to perform these actions until:

  1. Neither the maximum nor the minimum threshold is breached in the amount of time specified in the Auto Scaling rule
  2. The minimum server limit specified in the Auto Scaling rule has been reached
  3. The Server Farm runs out of Cloud Servers that can be started/stopped