Start 14-day free trial

Multi-Cloud Commitment Portfolio Engineering: RIs, Savings Plans, and CUDs Demystified

Master multi-cloud commitment management. Learn how to engineer Savings Plans, RIs, and CUDs without over-purchasing using CostGlass flat-rate FinOps.

Multi-Cloud Commitment Portfolio Engineering: RIs, Savings Plans, and CUDs Demystified
Figure 1: Architectural overview of Multi-Cloud Commitment Portfolio Engineering: RIs, Savings Plans, and CUDs Demystified.

Multi-cloud commitment portfolio engineering—balancing AWS Savings Plans, Reserved Instances (RIs), Azure Reservations, and Google Cloud Committed Use Discounts (CUDs)—requires strict mathematical alignment. Over-purchasing commitments while executing concurrent rightsizing or instance scheduling is a multi-million-dollar trap. To secure maximum coverage without waste, organizations must leverage evidence-based simulator modeling and deterministic overlap de-duplication, ensuring every rate-reduction strategy aligns with real-time operational shifts.

The Multi-Cloud Commitment Landscape: AWS, Azure, and GCP

Rate optimization is one of the most powerful levers in cloud financial management. However, managing commitments across a heterogeneous multi-cloud footprint introduces severe operational complexity. Each hyperscaler operates on different mechanics, billing granularities, and flexibility rules.

1. Amazon Web Services (AWS)

AWS offers a tiered hierarchy of commitments that trade flexibility for deeper discounts:

  • Compute Savings Plans (CSPs): Apply automatically to compute usage across Amazon EC2, AWS Fargate, and AWS Lambda, regardless of instance family, size, AZ, region, or operating system. They offer up to 66% savings.
  • EC2 Instance Savings Plans: Require commitment to a specific instance family within a selected region (e.g., m6g in us-east-1). They offer higher discounts (up to 72%) but restrict flexibility across families.
  • Reserved Instances (RIs): Standard and Convertible RIs apply to EC2, RDS, ElastiCache, OpenSearch, and Redshift. Standard RIs can be traded on the AWS RI Marketplace, whereas Convertible RIs allow changing instance attributes but require manual exchange actions.

2. Microsoft Azure

Azure unifies its commitment portfolio through the Azure Billing Profile Reader, allowing organizations to leverage:

  • Azure Reservations: Pre-purchased resource capacity for Virtual Machines, SQL Databases, Azure Cosmos DB, and storage. Reservations are scoped to a single resource group, subscription, or shared across an entire billing profile.
  • Azure Savings Plans for Compute: A flexible spend-based commitment model (similar to AWS Compute Savings Plans) that applies dynamically to Azure VMs, Dedicated Hosts, Container Instances, and Premium Functions.

3. Google Cloud Platform (GCP)

GCP operates primarily on Committed Use Discounts (CUDs):

  • Resource-Based CUDs: Tied to specific machine types (e.g., n2 or n2d) within a specific region. Highly restrictive but offering up to 70% discounts.
  • Spend-Based CUDs: Flexible commitments linked to dollar-per-hour spend across managed services like Google Cloud SQL, Google Kubernetes Engine (GKE) Autopilot, Cloud Run, and BigQuery.

The Over-Purchasing Trap: Why Concurrent Rightsizing and Scheduling Destroys Commitment ROI

The most common failure mode in enterprise FinOps is the lack of coordination between rate optimization (buying commitments) and usage optimization (rightsizing and scheduling).

Consider this mathematical scenario: Your current unoptimized AWS compute baseline is $100/hour. You purchase a 3-year No Upfront Compute Savings Plan for $70/hour, expecting a comfortable 70% coverage ratio.

Two weeks later, your platform engineering team executes two highly effective initiatives:

  1. They implement the CostGlass Multi-Cloud Instance Scheduler to turn off dev and staging environments on weekends and off-hours, cutting staging compute costs by 65%.
  2. They execute evidence-based rightsizing using p95 CloudWatch metrics, migrating over-provisioned m5.2xlarge instances to modern m6g.large shapes, slashing production compute requirements by another 30%.

Suddenly, your real-time compute baseline drops from $100/hour to $45/hour.

Because your Compute Savings Plan is a non-refundable, contractually binding commitment to pay $70/hour, you are now paying a $25/hour penalty for idle, unused commitment. Your effective savings rate has gone negative, and you are actively burning $18,000 every month on phantom compute.

[Unoptimized Baseline: $100/hr] 
       | 
       +---> Buy $70/hr CSP (70% Coverage) 
       | 
[Apply Rightsizing & Scheduling] 
       | 
       v 
[Real-Time Compute Drops to: $45/hr] 
       | 
       +---> WASTED COMMITMENT PENALTY: $25/hr ($18,000/mo) paid for zero active resources

To prevent this, CostGlass utilizes a deterministic overlap de-duplication engine (combine_overlapping). When analyzing billing telemetry from our 13 native connectors, CostGlass models rightsizing targets and active off-hours schedules before generating commitment recommendations. This ensures your baseline is calculated against your optimized future state, protecting you from over-purchasing contracts.


Cloud TCO Comparison: CostGlass vs. Competitors on Rate Optimization

Legacy FinOps tools penalize your engineering efficiency by tying their software fees directly to your cloud spend or, worse, taking a direct cut of your achieved savings.

  • nOps Alternatives: Competitors like nOps charge a percentage share of your rate-optimization savings. If their platform helps you save $50,000/month through automated commitments, they tax your efficiency by demanding thousands of dollars of those savings every month.
  • CloudZero and Finout Alternatives: CloudZero relies on opaque, annual-only contract structures that lock you in, while Finout charges heavy add-on surcharges for Kubernetes allocation and cost-per-customer unit economics.
  • The CostGlass Model: CostGlass operates on a transparent, predictable flat monthly fee: Free ($0 up to %5,000/mo spend), Starter ($99/mo), Growth ($499/mo), and Scale ($1,499/mo up to %100,000/mo spend). On our Scale plan, you keep 100% of every dollar saved—plus you get 8 dedicated senior DevOps/FinOps engineer-hours every month to safely execute these optimizations for you.
Feature / Capability CostGlass (Scale Plan) nOps CloudZero Vantage
Pricing Model Flat %1,499/mo (No spend tax) Percentage of Savings Tax Percentage of Spend Tiered Spend Brackets
SaaS/AI Connectors 13 Native (inc. OpenAI, Datadog) Limited Cloud Only Limited Cloud Only Standard Cloud Only
Commitment Simulator Included (What-If modeling) Included Included Included
Overlap De-duplication Yes (combine_overlapping) No No No
Hands-on Execution Yes (8 SRE hours/mo included) Software-Only Software-Only Software-Only
Data Sovereignty UK, EU, US, APAC Cells + BYOC US-Only SaaS US-Only SaaS US-Only SaaS

Build vs. Buy: The High Cost of DIY Commitment Simulation Pipelines

Many engineering organizations attempt to build their own internal commitment tracking spreadsheets or custom Python scripts to parse AWS CUR 2.0 or Azure MCA billing exports.

This DIY approach is a massive engineering and maintenance trap. Hyperscaler billing schemas are fragile and update continuously. For instance, reconciling AWS SavingsPlanNegation and DiscountedUsage columns alongside Azure MCA amortized records requires writing thousands of lines of complex ETL code. Your SRE team will spend dozens of hours every month maintaining a fragile internal dashboard that breaks whenever a cloud provider updates their billing API.

Instead of wasting $100,000+ in engineering payroll on internal billing pipelines, platform teams use CostGlass's 5-minute read-only setup. Here is an example of the clean, least-privilege Terraform configuration used to connect an AWS account to CostGlass's secure ingestion engine via an IAM Role with an ExternalId constraint:

# Terraform template to securely connect AWS Billing Telemetry to CostGlass

data "aws_caller_identity" "current" {}

resource "aws_iam_role" "costglass_readonly" {
  name               = "CostGlassReadOnlyIntegrationRole"
  assume_role_policy = data.aws_iam_policy_document.costglass_assume_role.json
}

data "aws_iam_policy_document" "costglass_assume_role" {
  statement {
    actions = ["sts:AssumeRole"]
    principals {
      type        = "AWS"
      identifiers = ["arn:aws:iam::471112253501:root"] # CostGlass secure ingestion principal
    }
    condition {
      test     = "StringEquals"
      variable = "sts:ExternalId"
      values   = ["cg_org_external_id_prod_99827"] # Your unique tenant external ID
    }
  }
}

resource "aws_iam_role_policy_attachment" "readonly_access" {
  role       = aws_iam_role.costglass_readonly.name
  policy_arn = "arn:aws:iam::aws:policy/ReadOnlyAccess"
}

This secure, read-only IAM integration allows CostGlass to safely pull your normalized multi-cloud billing ledger every 6 hours, feeding the data directly into our AI-driven What-If Commitment Simulator.


How to Configure and Automate Commitment Planning in CostGlass

Optimizing your commitment portfolio within CostGlass takes less than 15 minutes:

  1. Connect Your Clouds: Navigate to #settings:manage-accounts and follow our onboarding guides for AWS, Azure, and GCP.
  2. Verify Data Normalization: Ensure your multi-cloud billing records have compiled successfully. CostGlass unifies your cross-cloud spend into a standardized multi-cloud schema updated every 6 hours.
  3. Open the Commitment Portfolio: Navigate to the Commitments Dashboard (#finops:commitments) to view your current global coverage, utilization rates, and monthly waste.
  4. Run the What-If Simulator: Model 1-year vs. 3-year commitment horizons. Adjust the toggles to see the exact break-even month for No Upfront, Partial Upfront, and All Upfront scenarios.
  5. Coordinate SRE Execution: If you are on the Scale or Enterprise plan, do not buy these commitments alone. Open a ticket in the platform, and your dedicated CostGlass FinOps engineer will write the Terraform PRs, schedule the purchases, and verify the coverage post-buy.

Frequently Asked Questions

1. How does CostGlass calculate commitment recommendations if we are already rightsizing?

CostGlass calculates rightsizing and instance scheduling potential first. Our recommendation engine dynamically subtracts projected usage drops from your historical baseline, ensuring we only recommend commitments for your optimized, future-state compute needs.

2. Can CostGlass track commitments for Snowflake and Datadog?

Yes. CostGlass integrates natively with Snowflake (ORGANIZATION_USAGE key-pair integration) to track warehouse credit burn and capacity contracts, as well as Datadog log ingestion and custom metric commitments, unifying all SaaS and cloud spend into one pane of glass.

3. Is our commitment data safe? Do you store sensitive financial credentials?

No. CostGlass connects using read-only IAM roles with strict external ID validation. For highly regulated enterprises, CostGlass offers isolated regional cells in the US, UK, EU, and APAC, or a private in-account AWS BYOC (Bring Your Own Cloud) deployment so no billing metadata ever leaves your security boundary.

To eliminate cloud waste and build an optimized rate portfolio without paying a percentage of your savings as a tax, read our comprehensive CostGlass Savings Playbook or explore our flat-rate tiers on the CostGlass Pricing Page.

Ready to unify your multi-cloud costs?

See your AWS, Azure, GCP, and Kubernetes spend in canonical FOCUS 1.0 in under 15 minutes.

Start 14-day free trial →

More from the blog

View all →