Using PCF Autoscaler

What is the PCF autoscaler? 

A PCF service which scales to multiple instances based on app usage in the environment. 

How would you use it? 

Attributes that determine autoscaling: 

• CPU usage. 
• Throughput
• Latency

CPU Utilization

• Average CPU percentage for all instances of the app. 

Throughput:

• Total http requests per second (divided by total app instances). 
• No. Of bytes of data that you can move in the network over a given amount of time. 
• A fatter pipe or no. of lanes will help you move more people. 
• Speed test: through put. 
• Download: 50 MB
• Upload: 20 MB

Latency:

• Average latency of applications response to http requests. 
• Does not include router processing or network latency. 
• How long the smallest packet of data it takes to send from point A to point B. 

What can you configure? 

• Set minimum and maximum instance limits.
• Schedule these limits: reduce on weekends, scale up during peak holidays etc., 
• Scaling rules:
• CPU Utilization: 
• Scale up or down depending upon CPU usage in percentage
• HTTP throughput: 
• Scale up or down when the http through put is less than or greater than __ requests per second. 
• HTTP Latency: 
• Scale up or down if latency is less than or greater than __ milliseconds. 

Autoscaling schedules

There is an option to specify min and max limits and schedule autoscale limits based on schedules. This allows you to select the date time and repeatable on specific weekdays.

Note:

• If you manually scale an app, the autoscaler instance unbinds from that app and goes to manual setting.