Achieving Advanced Auto Scaling with AND/OR Logic in AWS Using Amazon CloudWatch Math Expressions

Overview

Auto Scaling is a critical AWS feature for maintaining optimal application performance by dynamically adjusting resource allocation based on demand. However, native AWS Auto Scaling Groups (ASGs) lack support for logical operators like AND and OR when defining scaling policies. This limitation poses challenges when scaling logic depends on multiple metrics, such as CPU and memory utilization.

Pioneers in Cloud Consulting & Migration Services

Reduced infrastructural costs
Accelerated application deployment

Get Started

Introduction

This blog introduces a solution using Amazon CloudWatch Math Expressions to implement advanced scaling logic. The approach utilizes AWS Auto Scaling, Amazon CloudWatch Alarms, and AWS CloudFormation, offering precise control over scaling actions, improved performance, and cost optimization.

AWS Services Used

Amazon Auto Scaling: Dynamically adjusts the number of Amazon EC2 instances in an ASG.
Amazon CloudWatch: Monitors metrics like CPU and memory utilization and triggers alarms.
CloudWatch Math Expressions: Enables custom metric calculations to implement complex logical conditions.
AWS CloudFormation: Automates resource creation and configuration as infrastructure-as-code.

The Problem: Lack of Logical Expressions in Auto Scaling

AWS Auto Scaling Groups use Amazon CloudWatch alarms to trigger scaling actions based on individual metrics. However, they don’t support logical conditions such as:

Scale-Out: Triggered if either CPU or memory usage exceeds thresholds.
Scale-In: Triggered only if both CPU and memory usage drop below thresholds.

Without native logical expressions, meeting such requirements requires workarounds.

The Solution: Amazon CloudWatch Math Expressions

Amazon CloudWatch Math Expressions allow creating custom metrics by applying mathematical and logical operations on existing metrics. This capability enables complex scaling policies by implementing AND/OR conditions.

Implementation

Define Thresholds:
- Scale-Out: CPU > 75% OR Memory > 70%
- Scale-In: CPU < 45% AND Memory < 50%
Create Math Expressions:
- OR Expression for Scale-Out:

scss

Copy code

OR_ScaleOut = IF((CPU_Utilization > 75) OR (Memory_Utilization > 70), 1, 0)

- AND Expression for Scale-In:

scss

Copy code

AND_ScaleIn = IF((CPU_Utilization < 45) AND (Memory_Utilization < 50), 1, 0)

3. Set Up Amazon CloudWatch Alarms:

- Scale-Out Alarm: Triggered when OR_ScaleOut breaches its threshold.
- Scale-In Alarm: Triggered when AND_ScaleIn breaches its threshold.

4. Attach Policies to ASG:

- Link the Scale-Out alarm to the scale-out policy to add instances.
- Link the Scale-In alarm to the scale-in policy to remove instances.

Automating with AWS CloudFormation

The following AWS CloudFormation snippet defines Amazon CloudWatch alarms with math expressions for scaling:

Resources:
  ScaleOutAlarm:
    Type: AWS::CloudWatch::Alarm
    Properties:
      AlarmName: "ScaleOutAlarm"
      ComparisonOperator: GreaterThanThreshold
      EvaluationPeriods: 1
      Threshold: 1
      Metrics:
        - Id: "cpu"
          MetricStat:
            Metric:
              Namespace: "AWS/EC2"
              MetricName: "CPUUtilization"
            Period: 60
            Stat: "Average"
        - Id: "memory"
          MetricStat:
            Metric:
              Namespace: "CWAgent"
              MetricName: "MemoryUtilization"
            Period: 60
            Stat: "Average"
        - Id: "OR_ScaleOut"
          Expression: "IF(cpu > 75 OR memory > 70, 1, 0)"
          Label: "Scale Out OR Condition"
      AlarmActions:
        - Ref: ScaleOutPolicy

  ScaleInAlarm:
    Type: AWS::CloudWatch::Alarm
    Properties:
      AlarmName: "ScaleInAlarm"
      ComparisonOperator: GreaterThanThreshold
      EvaluationPeriods: 1
      Threshold: 1
      Metrics:
        - Id: "cpu"
          MetricStat:
            Metric:
              Namespace: "AWS/EC2"
              MetricName: "CPUUtilization"
            Period: 60
            Stat: "Average"
        - Id: "memory"
          MetricStat:
            Metric:
              Namespace: "CWAgent"
              MetricName: "MemoryUtilization"
            Period: 60
            Stat: "Average"
        - Id: "AND_ScaleIn"
          Expression: "IF(cpu < 45 AND memory < 50, 1, 0)"
          Label: "Scale In AND Condition"
      AlarmActions:
        - Ref: ScaleInPolicy

Resources:

ScaleOutAlarm:

Type: AWS::CloudWatch::Alarm

Properties:

AlarmName: "ScaleOutAlarm"

ComparisonOperator: GreaterThanThreshold

EvaluationPeriods: 1

Threshold: 1

Metrics:

- Id: "cpu"

MetricStat:

Metric:

Namespace: "AWS/EC2"

MetricName: "CPUUtilization"

Period: 60

Stat: "Average"

- Id: "memory"

MetricStat:

Metric:

Namespace: "CWAgent"

MetricName: "MemoryUtilization"

Period: 60

Stat: "Average"

- Id: "OR_ScaleOut"

Expression: "IF(cpu > 75 OR memory > 70, 1, 0)"

Label: "Scale Out OR Condition"

AlarmActions:

- Ref: ScaleOutPolicy

ScaleInAlarm:

Type: AWS::CloudWatch::Alarm

Properties:

AlarmName: "ScaleInAlarm"

ComparisonOperator: GreaterThanThreshold

EvaluationPeriods: 1

Threshold: 1

Metrics:

- Id: "cpu"

MetricStat:

Metric:

Namespace: "AWS/EC2"

MetricName: "CPUUtilization"

Period: 60

Stat: "Average"

- Id: "memory"

MetricStat:

Metric:

Namespace: "CWAgent"

MetricName: "MemoryUtilization"

Period: 60

Stat: "Average"

- Id: "AND_ScaleIn"

Expression: "IF(cpu < 45 AND memory < 50, 1, 0)"

Label: "Scale In AND Condition"

AlarmActions:

- Ref: ScaleInPolicy

Benefits of This Approach

Enhanced Control: Precisely control scaling actions with custom logical conditions.
Resource Optimization: Avoid unnecessary scale-in or scale-out actions.
Automation: Use AWS CloudFormation to simplify deployment and management.
Cost Efficiency: Fine-tuned scaling policies reduce resource wastage and associated costs.

Conclusion

Amazon CloudWatch Math Expressions bridge the gap in logical scaling conditions for Auto Scaling Groups. By combining metrics with AND/OR logic, you can create robust, responsive scaling policies that optimize resource usage and application performance. Using AWS CloudFormation to automate these configurations ensures consistency and reduces manual effort.

Drop a query if you have any questions regarding Amazon CloudWatch Math Expressions and we will get back to you quickly.

Making IT Networks Enterprise-ready – Cloud Management Services

Accelerated cloud migration
End-to-end view of the cloud environment

Get Started

About CloudThat

CloudThat is a leading provider of Cloud Training and Consulting services with a global presence in India, the USA, Asia, Europe, and Africa. Specializing in AWS, Microsoft Azure, GCP, VMware, Databricks, and more, the company serves mid-market and enterprise clients, offering comprehensive expertise in Cloud Migration, Data Platforms, DevOps, IoT, AI/ML, and more.

FAQs

1. Can I use more than two metrics in my Amazon CloudWatch math expressions for scaling?

ANS: – Yes, you can use multiple metrics within a single Amazon CloudWatch math expression. Extending the math expression can incorporate additional metrics and more complex logical operations. For example, if you want to include a network throughput metric, CPU, and memory, you can add it to your expression to further refine your scaling logic. Remember that as you add more metrics, you must set thresholds carefully to avoid triggering scaling actions too frequently.

2. Will using Amazon CloudWatch math expressions for scaling increase my AWS costs?

ANS: – Using Amazon CloudWatch math expressions and custom alarms can incur additional charges, especially if you create multiple alarms or monitor high-resolution data. However, these costs are generally minor compared to the savings from optimized resource scaling. To manage costs, consider balancing the granularity of the metrics you track with the scaling responsiveness your application needs. AWS’s pricing page for Amazon CloudWatch provides more details on the specific costs of using metric math and alarms.

3. How do I know if my scale-out and scale-in policies work as intended?

ANS: – Monitor the scaling behavior closely after configuring your Amazon CloudWatch alarms with math expressions. You can use Amazon CloudWatch dashboards to visualize the alarm states and ASG activity logs to track scale-in and scale-out events. If you notice unexpected scaling behavior, adjust your threshold values or experiment with different logical combinations in your expressions. Regular testing and monitoring will help you fine-tune these policies for optimal results in real-world conditions.

WRITTEN BY Deepak S

Deepak S works as a Research Intern at CloudThat. His expertise lies in AWS's services. Deepak is good at haunting new technologies and automobile enthusiasts.