Key Features & Capabilities

This document provides a comprehensive walkthrough of AWS Tuner features, demonstrating how each capability addresses specific cloud cost optimization challenges through intelligent recommendations, automated scheduling, and spot instance management.

---

Table of Contents

1. Intelligent Recommendations
2. Automated Resource Scheduling
3. Spot Instance Management (SpotBot)
4. Multi-Account Optimization
5. Browser Extension Integration
6. Automated Remediation

---

Intelligent Recommendations

Recommendations Dashboard

Purpose: Provides actionable, prioritized cost optimization recommendations across your AWS infrastructure, powered by a Drools rule engine analyzing resource utilization and pricing data.

Key Features:

Category-Based Recommendations

AWS Tuner organizes recommendations into four strategic categories:

1. Cleaner - Remove unused and idle resources
2. OverProvisioned - Rightsize over-allocated resources
3. Modernization - Upgrade to cost-effective alternatives
4. Potential Benefits - Reserved Instances & Savings Plans

Example Savings (from demo):

Cleaner:          $48,645/month
OverProvisioned:  $28,087/month
Modernization:    $21,687/month
───────────────────────────────
Total Potential:  $98,419/month

Value: Immediately identify high-impact optimization opportunities organized by action type, enabling teams to prioritize based on effort vs. savings.

---

Cleaner Recommendations

Purpose: Identify and remove unused resources that continue to incur costs without providing value.

Covered Resources:

Resource Type	Demo Savings	Recommendation Count	Example Actions
EBS Snapshots	$11,278	20,292	Delete old snapshots no longer needed
NAT Gateway	$4,696	119	Remove idle NAT Gateways with no traffic
VPC Endpoint	$648	90	Delete unused VPC endpoints
S3 Multipart	$356	86	Clean up incomplete multipart uploads
DynamoDB	$366	162	Delete or archive idle tables

How It Works:

1. Resource Discovery: Scheduled jobs sync AWS resource metadata to MongoDB
2. Usage Analysis: CloudWatch metrics analyzed for activity (typically 30-day lookback)
3. Idle Detection: Drools rules evaluate usage thresholds
4. Recommendation Generation: Cost impact calculated and prioritized
5. Action Tracking: Track snoozed, implemented, or dismissed recommendations

Business Impact:

• Finance Teams: Reduce waste spending on forgotten resources
• Engineering Teams: Clean up technical debt and improve hygiene
• Compliance Teams: Reduce attack surface by removing unused infrastructure

---

OverProvisioned Recommendations

Purpose: Identify resources running with excessive capacity relative to actual utilization, enabling rightsizing for cost savings without performance impact.

Covered Resources:

Resource Type	Demo Savings	Recommendation Count	Optimization Strategy
EC2 Instances	$6,949	916	Downsize to optimal instance type
RDS Databases	$965	19	Reduce instance class based on CPU/memory
Redshift	$720	4	Rightsize cluster nodes

How It Works:

1. Metrics Collection: 30 days of CloudWatch metrics (CPU, memory, network, disk)
2. Utilization Analysis: Average utilization calculated across observation period
3. Threshold Evaluation: Drools rules check if utilization < 30% (configurable)
4. Pricing Optimization: AWS Pricing API consulted for optimal alternative instance types
5. Cost Comparison: Savings calculated only if new cost < current cost and savings > $0.005/month
6. Recommendation: Downsize action proposed with specific target instance type

Example Recommendation Details:

Resource: i-0abcd1234efgh5678 (m5.2xlarge)
Current Cost: $280.32/month
CPU Utilization: 12% (Max: 18%)
Memory Utilization: 25%

Recommended: m5.large
Recommended Cost: $70.08/month
Monthly Savings: $210.24
Annual Savings: $2,522.88

Action: Downsize EC2 instance type from m5.2xlarge to m5.large

Safety Features:

• Only recommends downsizing if current type != recommended type
• Respects instance architecture (x86 vs ARM)
• Considers tenancy (shared vs dedicated)
• Preserves platform details (Linux vs Windows licensing)
• Minimum savings threshold prevents noise

---

Modernization Recommendations

Purpose: Identify opportunities to upgrade AWS services to newer, more cost-effective versions or alternatives.

Covered Services:

Service	Modernization Type	Cost Impact	Example
RDS Extended Support	Upgrade to supported version	Avoid 30-50% surcharge	Upgrade MySQL 5.7 → 8.0
EKS Extended Support	Upgrade Kubernetes version	Avoid extended support fees	Upgrade EKS 1.23 → 1.29
ElastiCache	Migrate Redis → Valkey	Open-source alternative	Save on licensing
OpenSearch	EBS volume optimization	Upgrade to gp3 from gp2	20% cheaper, better performance
MS SQL Server	License optimization	Use License Mobility	Bring your own license

How It Works:

1. Version Detection: Identify service versions nearing end-of-support
2. Cost Analysis: Calculate extended support surcharges (RDS: +50% after standard support)
3. Alternative Evaluation: Identify equivalent open-source or optimized alternatives
4. Migration Path: Provide upgrade recommendations with compatibility checks
5. Savings Calculation: Compare current cost vs. modernized cost

Business Value:

• Avoid Surprise Costs: Extended support fees can increase costs by 30-50%
• Improved Security: Newer versions have latest security patches
• Better Performance: Modern versions often have performance improvements
• Open Source Savings: Migrate to open-source alternatives (e.g., Valkey vs Redis)

---

Potential Benefits Recommendations

Purpose: Identify opportunities for commitment-based discounts through Reserved Instances (RIs) and Savings Plans.

Covered Discount Programs:

1. EC2 Reserved Instances

   • 1-year or 3-year commitment
   • Up to 72% savings vs. On-Demand
   • Instance type specific

2. RDS Reserved Instances

   • Database engine specific
   • Up to 69% savings
   • Regional or multi-AZ

3. Compute Savings Plans

   • Flexible across instance families
   • Up to 66% savings
   • Apply to EC2, Fargate, Lambda

4. EC2 Instance Savings Plans

   • Specific instance family
   • Up to 72% savings
   • Most restrictive but highest discount

How It Works:

1. Usage Analysis: Analyze 30-90 days of stable workload patterns
2. Commitment Sizing: Calculate optimal hourly commitment
3. Discount Calculation: Compare On-Demand cost vs. RI/SP cost
4. Payback Period: Calculate break-even point
5. Recommendation: Suggest commitment level and term

Decision Matrix:

Workload Type	Recommended Program	Discount	Flexibility
Steady-state production	3-year RIs	72%	Low
Predictable growth	1-year Compute SP	66%	Medium
Variable instance types	Compute SP	66%	High
Short-term projects	Spot Instances	90%	Very High (interruptible)

---

Automated Resource Scheduling

Scheduler Dashboard

Purpose: Automate start/stop schedules for non-production resources to eliminate waste during off-hours.

Key Features:

Schedule Types

1. Account-Level Scheduling

   • Apply schedule to all resources in an AWS account
   • Use case: Dev/test environments shut down nights and weekends

2. Resource-Level Scheduling

   • Schedule specific resources (by resource ID)
   • Use case: Critical dev database stays running, others stop

3. Tag-Based Scheduling

   • Schedule resources matching AWS tag rules
   • Use case: All resources tagged Environment=Dev follow dev schedule

Schedule Configuration

Cron Expression Support:

# Stop at 7 PM weekdays
0 19 * * 1-5

# Start at 8 AM weekdays
0 8 * * 1-5

# Stop Friday 6 PM, Start Monday 8 AM
0 18 * * 5 (stop)
0 8 * * 1 (start)

Time Window Configuration:

• Start time (e.g., 8:00 AM)
• Stop time (e.g., 7:00 PM)
• Timezone support
• Exclude weekends/holidays

Supported Resources

Resource Type	Start Action	Stop Action	Savings Potential
EC2 Instances	StartInstances	StopInstances	65% (13hrs/day off)
RDS Databases	StartDBInstance	StopDBInstance	65% (13hrs/day off)
Auto Scaling Groups	Set desired capacity	Set capacity to 0	Variable

Savings Calculation:

Tuner tracks actual stop/start events and calculates realized savings:

Dev Environment Schedule:
- 10 m5.large EC2 instances ($0.096/hr each)
- Schedule: Stop 7 PM, Start 8 AM (13 hours off)
- Weekdays only (5 days/week)

Monthly Savings:
10 instances × $0.096/hr × 13 hrs/day × 22 days/month = $275/month per instance
Total: $2,750/month for 10 instances

Safety Features:

• Retry Mechanism: Failed stop/start operations automatically retried
• Event Logging: All schedule actions logged to QuartzEventLogs
• Exclude Production: Tag-based rules prevent accidental production shutdowns
• Audit Trail: Complete history of scheduled actions and who configured them

---

Spot Instance Management (SpotBot)

SpotBot Dashboard

Purpose: Identify workloads suitable for AWS Spot Instances and track savings from spot usage.

Key Features:

Spot Suitability Analysis

Evaluates workloads for spot instance compatibility:

Suitable Workloads:

• Batch processing jobs
• CI/CD pipelines
• Data analytics
• Machine learning training
• Containerized applications (ECS, EKS)
• Stateless web applications with load balancers

Unsuitable Workloads:

• Databases requiring consistent uptime
• Long-running stateful processes
• Real-time applications
• Applications without graceful shutdown handling

Interruption Risk Assessment

• Low Risk: Interruption rate < 5%
• Medium Risk: Interruption rate 5-10%
• High Risk: Interruption rate > 10%

Spot Pricing Intelligence

• Real-time spot price tracking
• Historical price trends
• Spot vs. On-Demand price comparison
• Savings calculator

Example Savings:

Workload: Nightly ETL Job
Instance: m5.4xlarge (16 vCPU, 64 GB RAM)

On-Demand: $0.768/hour × 4 hours/night × 30 days = $92.16/month
Spot:      $0.230/hour × 4 hours/night × 30 days = $27.60/month
───────────────────────────────────────────────────────────
Monthly Savings: $64.56 (70% discount)
Annual Savings:  $774.72

Spot Instance Best Practices

Diversification Strategy:

• Use multiple instance types (m5, m5a, m5n, m6i)
• Spread across availability zones
• Use capacity-optimized allocation strategy

Graceful Handling:

• Implement EC2 instance termination notice monitoring
• Use spot instance interruption notices (2-minute warning)
• Implement checkpointing for long-running jobs
• Use Auto Scaling Groups with mixed instance policies

---

Multi-Account Optimization

Account Management

Purpose: Manage optimization across multiple AWS accounts from a single pane of glass.

Key Features:

Cross-Account Access

• IAM Role Assumption: Uses AssumeRole for secure cross-account access
• Read-Only Permissions: Tuner only requires read access (no write permissions to customer AWS)
• Encrypted Credentials: All account credentials encrypted at rest
• Audit Logging: All API calls logged for compliance

Account Onboarding

1. Create CloudFormation StackSet with Tuner IAM role
2. Grant Tuner access via AssumeRole policy
3. Provide External ID for security
4. Tuner validates access and begins discovery
5. Recommendations available within 24 hours

Account-Level Controls

• Enable/disable recommendations per account
• Configure lookback periods per account
• Set custom thresholds per account
• Schedule recommendation refresh frequency

---

Browser Extension Integration

Purpose: Surface Tuner recommendations directly in the AWS Console while browsing resources.

Key Features:

Contextual Recommendations

When viewing an AWS resource (EC2, RDS, S3):

• Inline recommendations appear in AWS Console
• Cost savings estimate displayed
• One-click implementation (where safe)
• Snooze or dismiss recommendations

Supported AWS Console Pages

• EC2 instance details
• RDS database details
• S3 bucket details
• (Additional services to be documented)

Authentication

• JWT-based authentication with CloudKeeper web app
• Secure token storage in browser
• Automatic token refresh

Developer Value:

• Shift-Left Optimization: Developers see cost impact while working
• Reduced Context Switching: No need to leave AWS Console
• Immediate Action: Implement recommendations in real-time
• Education: Developers learn cost optimization best practices

---

Automated Remediation

Purpose: Automatically implement low-risk recommendations without manual intervention.

Eligible Recommendations:

Recommendation Type	Auto-Remediation Support	Risk Level
EBS Snapshot Cleanup	✅ Yes	Low (no active attachments)
S3 Multipart Cleanup	✅ Yes	Low (incomplete uploads only)
Unused AMI Deletion	✅ Yes	Low (no running instances)
EC2 Rightsizing	❌ No (requires testing)	Medium-High
RDS Rightsizing	❌ No (requires testing)	High

Safety Features:

• Dry-Run Mode: Test remediation without applying changes
• Approval Workflow: Require approval for specific account/resource types
• Rollback Capability: Undo remediation if issues detected
• Audit Trail: Complete history of automated actions

Configuration:

auto-remediation:
  enabled: true
  risk-level: low  # only auto-remediate low-risk recommendations
  approval-required: false
  notification:
    email: ops-team@company.com
    slack: #cost-optimization

---

Integration Points

CloudKeeper Platform Integration

AWS Tuner integrates with other CloudKeeper modules:

1. AWS Lens (Shared Data Warehouse)

   • Cost & Usage Report (CUR) data from Snowflake
   • Historical cost trends
   • Service-level spending

2. AuthX (Authentication)

   • Single sign-on (SSO)
   • Role-based access control (RBAC)
   • Multi-factor authentication (MFA)

3. Config Server (Configuration Management)

   • Centralized configuration
   • Feature flags
   • Environment-specific settings

External Integrations

1. AWS Services

   • EC2 API (instance metadata, metrics)
   • RDS API (database details)
   • CloudWatch API (performance metrics)
   • Cost Explorer API (pricing data)
   • Pricing API (rate cards)

2. Message Queue (RabbitMQ)

   • Async event processing
   • Job scheduling
   • Recommendation pipeline

3. Databases

   • MongoDB (recommendations, user preferences)
   • MySQL (accounts, user management)
   • Snowflake (cost analytics)
   • Redis (caching)

---

Summary

AWS Tuner provides comprehensive cost optimization through:

• 42+ Recommendation Types across compute, storage, database, and networking
• 40+ Drools Rules for intelligent, context-aware recommendations
• Automated Scheduling for non-production resource management
• Spot Instance Intelligence for batch and interruptible workloads
• Multi-Account Management for enterprise-scale optimization
• Browser Extension for developer-friendly recommendations
• Automated Remediation for low-risk optimizations

Total Savings Potential (from demo): $98,419/month ($1,181,028/year)

Next Steps:

• Business Value & ROI
• Getting Started Guide
• Architecture Deep Dive