Reshaping

Learn more about how to reshape your infrastructure to benefit from savings and efficiency.

Reshaping

North's Reshaping feature uses advanced machine learning algorithms to continuously monitor your AWS infrastructure and deliver intelligent, actionable optimization recommendations for your resources. Our solution adapts to your unique workload patterns and automatically retrains itself based on the changes you implement and the real-world outcomes of those modifications.

How the ML-Driven Reshaping Works

Our machine learning models ingest a broad set of workload parameters collected from your infrastructure using real-time and historical log metrics. These models analyze usage trends, workload variability, and resource efficiency to identify where resizing, rightsizing, or other optimization actions will either reduce costs or enhance workload performance.

Reshaping is not a one-time action. As you accept and implement our recommendations—such as resizing an instance type or adjusting memory limits—our models observe the efficiency and cost outcomes, retraining automatically to provide even more accurate, personalized suggestions over time.

How the App Works

• Continuous Monitoring: Our intelligent system keeps a constant watch on your EC2 instances, ECS services, EBS volumes, Lambda functions, Auto Scaling groups, and more, collecting the necessary metrics for optimization.

• AI Recommendations List: The dashboard displays a curated list of AI-driven suggestions. You’ll see both recommendations to save money by downsizing where possible and guidance to upsize resources that are performance-constrained. Each recommendation comes with clear analytics and the rationale behind the choice—just click the AI Suggestion button for full details.

• Collaboration and Actions: Share any recommendation directly with your team via Slack, by creating a Jira ticket, or sending by email—all from within North’s platform. This streamlines decision-making and speeds up implementation.

Supported Resources & Optimization Types

Our ML-driven Reshaping logic covers:

• EC2 Instances: Right-size based on actual CPU/memory demand, balancing cost and performance; evaluates migration complexity and platform differences.

• ECS Services: Container task optimizations for Fargate/ECS, including memory/CPU tuning at service/task granularity.

• EBS Volumes: Volume type and IOPS tuning, flagging over-provisioned storage.

• Lambda Functions: Memory/provisioned concurrency rightsizing and invocation cost analysis.

• Auto Scaling Groups: Compute family and capacity optimizations, with recommended scaling strategies.

• Idle Resources: Automated identification of low-use or unused resources with clear cost impact and safe removal suggestions.

How Recommendations are Generated

1. Data Collection:

   • Real-time and historical log metrics are aggregated (32-day default lookback, configurable)

   • ML models draw from CloudWatch, Cost Explorer, and other sources

2. Deep Analysis:

   • Our algorithms apply multi-variate analysis of utilization (P95 CPU/memory usage, workload seasonality, anomaly detection, and more)

   • Dynamically adapts to spikes, troughs, and changing patterns

3. Recommendation Generation:

   • Each finding is rated (over-provisioned, under-provisioned, optimized)

   • Monthly savings or performance gains are estimated, migration effort assessed, and analytics provided for transparency

4. Learning & Retraining:

   • When you apply a suggestion, results are measured and fed back into the ML system

   • Models retrain continually, growing more accurate as your infrastructure and requirements evolve

5. Recommendation Storage & History:

   • All recommendations are stored, versioned, and tracked in DynamoDB with account and resource granularity

   • Historical effectiveness metrics available for review

Configuration

You can tailor the analysis through these preferences:

• Lookback Period: (14/32/93 days—longer lookbacks yield deeper insights)

• CPU/Memory Headroom: (set thresholds per your risk tolerance; defaults are conservative)

• CPU Utilization Percentile: (P90, P95, P99.5)

• Scope: Organization-wide, specific accounts, or individual resources

• Savings Estimation Mode: Incorporates existing reservations and real discounting for real-world savings projections

Interpreting and Using Recommendations

Each ML-driven suggestion gives you:

• Current vs. Recommended Configuration

• Monthly Savings or Performance Gains

• Migration Complexity

• Underlying Analytics (utilization history, risk headroom, anomaly notes)

• Change Tracking and Effectiveness

You can review, accept, ignore, or bulk apply suggestions, and track the realized effect inside North.