Skip to main content

Getting Started with AWS Tuner

This guide walks you through onboarding AWS Tuner, from initial account setup to implementing your first cost optimizations.

Table of Contents​

  1. Prerequisites
  2. Account Onboarding
  3. Initial Discovery & Analysis
  4. Understanding Recommendations
  5. Implementing Your First Optimizations
  6. Setting Up Scheduler
  7. Browser Extension Installation
  8. Best Practices

Prerequisites​

Access Requirements​

  • CloudKeeper Account: Active CloudKeeper platform account
  • AWS Account Access: AWS accounts to be optimized
  • IAM Permissions: Ability to create IAM roles in target AWS accounts
  • User Role: Minimum role of "Tuner User" in CloudKeeper platform

Technical Requirements​

  • Browser: Chrome, Firefox, Edge, or Safari (latest version)
  • Network Access: HTTPS connectivity to CloudKeeper platform
  • AWS Console Access: For implementing recommendations
  • Basic AWS service familiarity (EC2, RDS, S3)
  • Understanding of AWS pricing models (On-Demand, Reserved, Spot)
  • AWS IAM role and policy concepts

Account Onboarding​

Step 1: Access AWS Tuner​

  1. Log in to CloudKeeper platform at https://app.cloudkeeper.com
  2. Click "Switch to Tuner" in the left navigation
  3. You'll be redirected to the Tuner dashboard

Tuner Dashboard

Option A: CloudFormation StackSet (Recommended for Multiple Accounts)

# 1. Download CloudFormation template
wget https://cloudkeeper-public.s3.amazonaws.com/tuner/cloudkeeper-tuner-role.yaml

# 2. Create stack in AWS CloudFormation
aws cloudformation create-stack \
  --stack-name CloudKeeperTunerRole \
  --template-body file://cloudkeeper-tuner-role.yaml \
  --parameters \
    ParameterKey=ExternalId,ParameterValue=YOUR_EXTERNAL_ID \
    ParameterKey=CloudKeeperAccountId,ParameterValue=123456789012 \
  --capabilities CAPABILITY_NAMED_IAM

Option B: Manual IAM Role Creation

  1. Navigate to Tuner → Accounts → Link Account
  2. Click "Generate IAM Policy"
  3. Copy the generated IAM policy JSON
  4. In AWS Console → IAM → Roles → Create Role
  5. Select "Another AWS Account"
  6. Enter CloudKeeper Account ID: 123456789012
  7. Check "Require external ID" and enter your unique External ID
  8. Attach the generated policy
  9. Name role: CloudKeeperTunerRole
  10. Copy the Role ARN back to CloudKeeper Tuner

IAM Policy (Read-Only):

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "ec2:Describe*",
        "rds:Describe*",
        "cloudwatch:GetMetricStatistics",
        "cloudwatch:ListMetrics",
        "pricing:GetProducts",
        "ce:GetCostAndUsage",
        "s3:ListBucket",
        "s3:GetBucketLocation",
        "dynamodb:DescribeTable",
        "dynamodb:ListTables",
        "redshift:Describe*",
        "elasticache:Describe*",
        "autoscaling:Describe*"
      ],
      "Resource": "*"
    }
  ]
}

Step 3: Verify Access​

  1. Click "Verify Access" in Tuner
  2. Tuner will test IAM role assumption
  3. Wait for green checkmark: ✅ "Account linked successfully"
  4. Initial resource discovery begins automatically

Troubleshooting:

ErrorCauseSolution
"AssumeRole failed"External ID mismatchVerify External ID matches exactly
"Access Denied"Insufficient permissionsReview IAM policy, ensure all actions allowed
"Role not found"Role ARN incorrectDouble-check Role ARN, ensure role exists

Initial Discovery & Analysis​

What Happens After Onboarding?​

Timeline:

Hour 0:    Account linked successfully
Hour 1:    Resource metadata sync begins (EC2, RDS, S3, etc.)
Hour 4-6:  CloudWatch metrics collection starts
Hour 24:   First recommendations available
Hour 48:   Full analysis complete (30-day lookback)

Monitoring Discovery Progress​

Navigate to Tuner → Accounts → [Your Account] → Discovery Status:

✅ EC2 Instances:        456 resources discovered
✅ RDS Databases:        23 resources discovered
⏳ CloudWatch Metrics:   Collecting (78% complete)
✅ EBS Volumes:          1,234 volumes discovered
✅ EBS Snapshots:        8,765 snapshots discovered
⏳ S3 Buckets:          Analyzing (45% complete)
✅ NAT Gateways:         12 gateways discovered
✅ VPC Endpoints:        34 endpoints discovered

First Look at Your Environment​

After 24 hours, check Tuner → Recommendations:

Recommendations Dashboard

What You'll See:

Total Potential Savings: $XX,XXX/month

Category Breakdown:
├─ Cleaner:          $XX,XXX/month  (remove unused resources)
├─ OverProvisioned:  $XX,XXX/month  (rightsize over-allocated)
├─ Modernization:    $XX,XXX/month  (upgrade services)
└─ Potential Benefits: $XX,XXX/month  (RIs & Savings Plans)

Understanding Recommendations​

Recommendation Anatomy​

Each recommendation contains:

  1. Category: Cleaner, OverProvisioned, Modernization, or Potential Benefits
  2. Resource Details: Instance ID, name, region, account
  3. Current State: Current configuration and cost
  4. Recommended Action: Specific change to implement
  5. Cost Impact: Monthly and annual savings
  6. Risk Level: Low, Medium, or High
  7. Metadata: Utilization metrics, sizing details

Example Recommendation:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📋 EC2 Instance Rightsizing

Resource ID:    i-0abcd1234efgh5678
Resource Name:  api-server-prod-3
Region:         us-east-1
Account:        Production (123456789012)

Category:       OverProvisioned
Risk Level:     Medium

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Current Configuration:
  Instance Type:    m5.2xlarge
  CPU:              8 vCPUs
  Memory:           32 GB
  Cost/Hour:        $0.384
  Cost/Month:       $280.32

Utilization (30-day average):
  CPU:              12% (max: 18%)
  Memory:           25% (max: 32%)
  Network:          Low

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Recommended Configuration:
  Instance Type:    m5.large
  CPU:              2 vCPUs
  Memory:           8 GB
  Cost/Hour:        $0.096
  Cost/Month:       $70.08

Savings:
  Monthly:          $210.24 (75% reduction)
  Annual:           $2,522.88

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Action Required:
  1. Test m5.large in staging environment
  2. Monitor performance for 1 week
  3. If acceptable, apply to production
  4. Schedule change during maintenance window

Recommended By: Drools Rule Engine (oper_provisioned_ec2_instance_rules.drl)
Confidence: High (based on 30 days data)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Prioritizing Recommendations​

Priority Matrix:

PriorityCriteriaTypical SavingsRisk
P0 - Critical> $5K/month, Low risk$5K-$50KLow
P1 - High> $1K/month, Low-Medium risk$1K-$5KLow-Medium
P2 - Medium> $100/month, Any risk$100-$1KVaries
P3 - Low< $100/month< $100Low

Recommended Approach:

  1. Week 1: Implement all P0 cleaner recommendations (low-risk wins)
  2. Week 2-3: Test P1 rightsizing recommendations in non-prod
  3. Week 4: Implement P1 rightsizing in production with monitoring
  4. Month 2: Address P2 and modernization recommendations

Implementing Your First Optimizations​

Quick Win: EBS Snapshot Cleanup​

Goal: Remove orphaned EBS snapshots to save $5K-$15K/month

Navigate to Tuner → Recommendations → Cleaner:

Cleaner Recommendations

Step-by-Step:

  1. Review Recommendations:

    • Click "EBS Snapshots" tile
    • Review list of snapshots recommended for deletion
    • Filter by age: "Older than 90 days"

  2. Validate Safety:

    # For each snapshot, verify no attached volumes
    aws ec2 describe-snapshots --snapshot-ids snap-0abcd1234 \
      --query 'Snapshots[0].VolumeId'

    # If Volume is deleted, snapshot is safe to remove

  3. Implement (Choose one):

    Option A: Automated Remediation (Recommended)

    • Check snapshots to delete
    • Click "Auto Remediate"
    • Confirm deletion
    • Tuner deletes snapshots automatically

    Option B: Manual Deletion

    • Export snapshot IDs to CSV
    • Use AWS CLI or Console to delete
    • Mark as "Implemented" in Tuner

  4. Monitor Results:

    • Savings appear in next month's bill
    • Tuner tracks realized savings vs. projected

Expected Results:

  • Time to Implement: 30 minutes - 2 hours
  • Monthly Savings: $5K-$15K (typical)
  • Risk: Very Low (only deletes unattached snapshots)

Medium Impact: EC2 Rightsizing​

Goal: Reduce EC2 costs by 20-40% through rightsizing

Navigate to Tuner → Recommendations → OverProvisioned:

OverProvisioned Recommendations

Step-by-Step:

  1. Select Test Candidate:

    • Start with non-production environment
    • Choose instance with > $100/month savings
    • Low utilization (< 20% CPU)

  2. Validate Recommendation:

    # Review CloudWatch metrics manually
    aws cloudwatch get-metric-statistics \
      --namespace AWS/EC2 \
      --metric-name CPUUtilization \
      --dimensions Name=InstanceId,Value=i-0abcd1234 \
      --start-time 2025-09-26T00:00:00Z \
      --end-time 2025-10-26T00:00:00Z \
      --period 3600 \
      --statistics Average,Maximum

  3. Test in Staging:

    • Create AMI of current instance
    • Launch new instance with recommended type
    • Run load tests
    • Monitor performance for 1 week

  4. Implement in Production:

    # 1. Stop instance
    aws ec2 stop-instances --instance-ids i-0abcd1234

    # 2. Modify instance type
    aws ec2 modify-instance-attribute \
      --instance-id i-0abcd1234 \
      --instance-type m5.large

    # 3. Start instance
    aws ec2 start-instances --instance-ids i-0abcd1234

    # 4. Monitor CloudWatch for 24 hours

  5. Mark as Implemented:

    • In Tuner, click "Mark as Implemented"
    • Tuner tracks actual savings vs. projected

Expected Results:

  • Time to Implement: 1-2 weeks (including testing)
  • Monthly Savings: $200-$500 per instance
  • Risk: Medium (requires performance validation)

Setting Up Scheduler​

Goal: Automatically stop non-production resources during off-hours

Navigate to Tuner → Scheduler:

Scheduler

Basic Schedule Configuration​

Use Case: Development environment used Monday-Friday, 8 AM - 7 PM

Step-by-Step:

  1. Create Schedule:

    • Click "Create Schedule"
    • Name: "Dev Environment Schedule"
    • Schedule Type: "Tag-Based"

  2. Define Time Windows:

    Start Time:  Monday-Friday 7:00 AM EST
    Stop Time:   Monday-Friday 7:00 PM EST
    Weekend:     Completely off
    Timezone:    America/New_York

  3. Configure Tag Rules:

    Match resources where:
    - Tag Key: "Environment"
    - Tag Value: "Dev"

    AND

    - Tag Key: "Scheduler"
    - Tag Value: "Enabled"

  4. Select Resource Types:

    • ✅ EC2 Instances
    • ✅ RDS Databases
    • ❌ Auto Scaling Groups (optional)

  5. Safety Settings:

    • ✅ Enable retry on failure
    • ✅ Send email alerts
    • ❌ Allow manual override (optional)

  6. Review & Activate:

    • Preview matched resources (e.g., 45 EC2, 12 RDS)
    • Review projected savings
    • Click "Activate Schedule"

Savings Calculation​

Tuner automatically calculates expected savings:

Development Environment:
├─ 45 EC2 instances (avg $0.10/hr)
├─ Off hours: 127 hours/week
├─ Monthly off hours: 548 hours
└─ Monthly savings: 45 × $0.10 × 548 = $2,466/month

12 RDS databases (avg $0.15/hr):
└─ Monthly savings: 12 × $0.15 × 548 = $986/month

Total Monthly Savings: $3,452
Annual Savings: $41,424

Monitoring Scheduler​

Dashboard Metrics:

  • Success Rate: 99.2% (446/450 operations)
  • Failed Stops: 2 (retry successful)
  • Failed Starts: 2 (manual intervention required)
  • Realized Savings: $3,387/month (98% of projected)

Browser Extension Installation​

Goal: See Tuner recommendations directly in AWS Console

Installation Steps​

  1. Chrome Web Store:

  2. Firefox Add-ons:

  3. Authentication:

    • Log in to CloudKeeper web app
    • Extension automatically captures JWT token
    • Green badge indicates: ✅ Authenticated

Using the Extension​

When viewing an AWS resource in AWS Console:

EC2 Instance Example:

┌─────────────────────────────────────────────────┐
│ 💡 CloudKeeper Tuner Recommendation             │
├─────────────────────────────────────────────────┤
│ Instance: i-0abcd1234efgh5678                   │
│ Current: m5.2xlarge ($280.32/month)             │
│ CPU Utilization: 12% (30-day avg)               │
│                                                  │
│ Recommended: m5.large ($70.08/month)            │
│ Monthly Savings: $210.24                        │
│                                                  │
│ [Implement] [Snooze 30 days] [Dismiss]         │
└─────────────────────────────────────────────────┘

Actions:

  • Implement: Creates Jira ticket or schedules change
  • Snooze: Hide recommendation for 30/60/90 days
  • Dismiss: Permanently dismiss (requires reason)

Best Practices​

For Maximum Savings​

  1. Start with Low-Risk Wins:

    • Cleaner recommendations first (snapshots, idle resources)
    • Build confidence before rightsizing production

  2. Use Phased Approach:

    • Week 1: Non-production optimizations
    • Week 2-3: Test production changes in staging
    • Week 4+: Implement in production with monitoring

  3. Enable Automated Remediation:

    • Only for low-risk recommendations
    • Review audit logs weekly
    • Disable if unexpected issues arise

  4. Schedule Non-Production Resources:

    • Immediate 40-60% savings on dev/staging
    • Zero performance impact on production
    • Tag resources properly for accuracy

For Safety​

  1. Always Test First:

    • Never apply rightsizing directly to production
    • Use staging environment mirror
    • Run load tests before production rollout

  2. Monitor After Changes:

    • CloudWatch alarms for CPU, memory, response time
    • Application performance monitoring (APM)
    • Rollback plan ready

  3. Document Implementations:

    • Track which recommendations implemented
    • Note any issues encountered
    • Share learnings with team

  4. Review Monthly:

    • New recommendations generated continuously
    • Workload patterns change over time
    • Adjust thresholds as needed

For Governance​

  1. Define Approval Workflows:

    Low-Risk (< $500/month):   Engineering team approval
    Medium-Risk ($500-$5K):    Engineering + Finance approval
    High-Risk (> $5K):         C-level approval required

  2. Tag Resources Properly:

    • Environment: Dev/Staging/Prod
    • Owner: Team responsible
    • CostCenter: For chargeback
    • Scheduler: Enable/Disable scheduling

  3. Audit Trail:

    • All changes logged in Tuner
    • Export monthly reports
    • Track realized vs. projected savings

Common Pitfalls to Avoid​

❌ Don't:

  • Implement all recommendations immediately without testing
  • Ignore utilization spikes (e.g., month-end batch jobs)
  • Disable scheduler without communication
  • Delete snapshots without verifying no restore needs

✅ Do:

  • Validate recommendations with your domain knowledge
  • Consider workload patterns (daily, weekly, seasonal)
  • Communicate schedule changes to developers
  • Keep critical snapshots (disaster recovery)

Success Metrics​

Track your optimization progress:

Month 1 (Discovery & Quick Wins):

  • ✅ All accounts onboarded
  • ✅ 100% snapshot cleanup completed
  • ✅ Scheduler enabled for dev/staging
  • 📊 Savings: 5-10% of AWS spend

Month 2 (Rightsizing Testing):

  • ✅ Non-production instances rightsized
  • ✅ Production test candidates identified
  • 🧪 Performance testing in progress
  • 📊 Savings: 10-15% of AWS spend

Month 3 (Production Rollout):

  • ✅ Production rightsizing completed
  • ✅ Modernization recommendations reviewed
  • ✅ Spot instances for batch workloads
  • 📊 Savings: 15-25% of AWS spend

Month 6 (Continuous Optimization):

  • ✅ Reserved Instances purchased
  • ✅ Automated remediation enabled
  • ✅ Browser extension adopted by 80% of engineers
  • 📊 Savings: 20-35% of AWS spend

Getting Help​

Support Channels​

Escalation Path​

  1. Level 1: Tuner UI help documentation and tooltips
  2. Level 2: Email support (24-hour response SLA)
  3. Level 3: Customer Success Manager (existing customers)
  4. Level 4: Solutions Architect (enterprise customers)

Training Resources​

Next Steps​

  1. Review Key Features: Key Features & Capabilities
  2. Understand ROI: Business Value & ROI Examples
  3. Deep Dive on Architecture: Architecture Overview

Ready to get started? Link your first AWS account →