Amazon EBS Cost Optimizer

Operationalize AWS Compute Optimizer EBS recommendations with real CloudWatch usage data, dollar-denominated impact, and safe automated remediation.

Overview

AWS Compute Optimizer surfaces over- and under-provisioned EBS volumes, but acting on those findings at scale is a manual, per-volume process today. This utility closes that loop end-to-end:

Pulls every Compute Optimizer EBS recommendation flagged NotOptimized.
Validates each recommendation against 30 days of actual workload data from CloudWatch (peak IOPS and throughput).
Calculates a right-sized target with a configurable safety buffer, floored at the gp3 baseline (3000 IOPS / 125 MiB/s) and capped at the volume type's service ceiling.
Estimates monthly and annual cost impact in USD using a configurable pricing model (defaults to AWS list prices).
Produces a CSV report sorted by attached EC2 instance with per-instance subtotals, an orphan-volume section, and a grand total.
With explicit opt-in, applies the changes via ec2:ModifyVolume (online, no detach, no I/O pause).

Why this is helpful

Compute Optimizer dashboards routinely show tens of thousands of dollars in flagged EBS savings, but customers struggle to act on them because:

Recommendations are heuristic. Customers want to verify against real workload peaks before reducing provisioned IOPS or throughput.
Per-volume console actions don't scale. Modifying 300 volumes manually is a multi-day exercise prone to errors.
Financial impact isn't visible. The dashboard shows aggregate dollars but not per-volume or per-instance numbers stakeholders can act on.
Orphan volumes are an adjacent problem. Unattached available volumes are pure waste and rarely get cleaned up.

This tool addresses all four in a single, auditable run.

Use cases

Cost optimization (downsize)

Identify gp3 volumes provisioned with IOPS or throughput well above their 30-day peak, generate a per-volume savings estimate, and apply right-sized values during a controlled change window.

python3 ebs_cost_optimizer.py --region us-east-1 --direction downsize

The CSV gives you monthly_delta_usd per volume, a SUBTOTAL per EC2 instance, and a GRAND_TOTAL row at the bottom. Hand it to FinOps for sign-off, then:

python3 ebs_cost_optimizer.py --region us-east-1 --direction downsize \
    --volume-ids-file approved.txt --apply

Performance remediation (upsize)

Surface volumes whose workload is being throttled by a low IOPS or throughput ceiling. The script flags these as MODIFY UPSIZE rows so app owners can react before users notice.

python3 ebs_cost_optimizer.py --region us-east-1 --direction upsize

Idle resource cleanup (orphans)

Inventory unattached available volumes with their age, encryption status, and the monthly cost they represent. The tool reports only — deletion stays a manual step, gated by your snapshot and change-management process.

python3 ebs_cost_optimizer.py --region us-east-1 --orphans-only --output orphans.csv

Multi-account governance

Wrap the script in an STS AssumeRole loop across organization member accounts to produce a consolidated report from the management account or a delegated admin. The script is per-account by design so the apply path stays auditable per account boundary.

How it works

Data flow

┌────────────────────────┐
│ AWS Compute Optimizer  │  GetEBSVolumeRecommendations (paginated)
└──────────┬─────────────┘
           │  NotOptimized findings
           ▼
┌────────────────────────┐
│ Amazon CloudWatch      │  GetMetricData (single call per volume,
└──────────┬─────────────┘    all 4 EBS metrics, 5-minute resolution)
           │  peak IOPS + throughput
           ▼
┌────────────────────────┐
│ Amazon EC2             │  DescribeVolumes, DescribeVolumesModifications
└──────────┬─────────────┘
           │  current Iops/Throughput, cooldown state
           ▼
┌────────────────────────┐
│ Right-sizing engine    │  target = max(floor, peak × (1 + buffer))
│ Cost engine            │           capped at type service limit
└──────────┬─────────────┘
           │
           ▼
┌────────────────────────┐
│ CSV report             │  sorted by attached_instances,
└──────────┬─────────────┘    subtotals + grand total
           │
           ▼  (with --apply)
┌────────────────────────┐
│ ec2:ModifyVolume       │  online change, no detach
└────────────────────────┘

Volume types

By default, the script focuses on gp3 since IOPS and throughput are independently tunable for cost optimization. Use --all-types to widen.

Type	IOPS tunable	Throughput tunable	Default behavior
gp3	yes	yes	full right-size
io1	yes	no	filtered out
io2	yes	no	filtered out
gp2	no (size-bound)	no	filtered out
st1/sc1	no	no	filtered out

Prerequisites

Python 3.8 or later
boto3 1.26+ (AWS SDK)
openpyxl 3.0+ — required only when generating .xlsx reports
AWS credentials with the IAM permissions listed below
AWS Compute Optimizer opted in for the target account (documentation)

Install the dependencies

Pick whichever fits your environment.

Amazon Linux 2023 (common on EC2):

sudo dnf install -y python3-boto3 python3-openpyxl

Amazon Linux 2:

sudo yum install -y python3-boto3 python3-openpyxl

Debian / Ubuntu:

sudo apt-get install -y python3-boto3 python3-openpyxl

RHEL / CentOS / Rocky:

sudo yum install -y python3-boto3 python3-openpyxl

macOS / generic Linux with pip:

pip3 install --user -r requirements.txt

If pip3 itself is missing on Amazon Linux:

sudo dnf install -y python3-pip   # AL2023
sudo yum install -y python3-pip   # AL2

openpyxl is optional. Skip it if you only want CSV output. The script will detect it's missing and emit a platform-specific install hint if you ask for .xlsx without it installed.

Required IAM permissions

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "compute-optimizer:GetEBSVolumeRecommendations",
        "ec2:DescribeVolumes",
        "ec2:DescribeVolumesModifications",
        "cloudwatch:GetMetricData",
        "sts:GetCallerIdentity"
      ],
      "Resource": "*"
    },
    {
      "Sid": "ApplyOnly",
      "Effect": "Allow",
      "Action": ["ec2:ModifyVolume"],
      "Resource": "arn:aws:ec2:*:*:volume/*"
    }
  ]
}

For read-only operation, omit the ApplyOnly statement.

Quickstart

git clone https://github.com/aws-samples/ebs-cost-optimizer.git
cd ebs-cost-optimizer
pip install -r requirements.txt

# Read-only report
python3 ebs_cost_optimizer.py --region us-east-1 --include-orphans

Common workflows

# Cost-savings only (drop upsize recommendations)
python3 ebs_cost_optimizer.py --region us-east-1 --direction downsize

# Performance fixes only
python3 ebs_cost_optimizer.py --region us-east-1 --direction upsize

# Override default pricing with customer billing rates
python3 ebs_cost_optimizer.py --region us-east-1 \
    --price-storage 0.075 --price-iops 0.0045 --price-throughput 0.036

# Apply to a vetted list (preferred apply path)
python3 ebs_cost_optimizer.py --region us-east-1 \
    --volume-ids-file approved.txt --apply

# Apply to every flagged volume in scope (must opt in)
python3 ebs_cost_optimizer.py --region us-east-1 --apply --apply-all

CLI reference

Flag	Default	Purpose
`--region`	required	AWS region
`--profile`	none	AWS CLI profile
`--days`	30	CloudWatch lookback (1-63)
`--buffer`	20	Headroom percentage above peak
`--min-iops`	3000	Floor for IOPS (gp3 baseline)
`--min-throughput`	125	Floor for throughput MiB/s
`--max-workers`	8	Parallel CloudWatch threads
`--output`	`EBS_Cost_Optimization.xlsx`	Report path. Use `.xlsx` for styled Excel (default), `.csv` for plain text
`--no-timestamp`	off (timestamps on)	Disable UTC timestamp suffix in filename
`--no-subtotals`	off (subtotals on)	Disable per-instance subtotal rows
`--pricing-file`	none	JSON file with gp3 prices
`--price-storage`	0.08	Override $/GiB-month
`--price-iops`	0.005	Override $/IOP-month above 3000
`--price-throughput`	0.04	Override $/MiB/s-month above 125
`--gp3-only`	on	Focus only on gp3 volumes
`--all-types`	off	Include io1/io2/gp2
`--direction`	`both`	`both` / `upsize` / `downsize`
`--include-orphans`	off	Append unattached volumes
`--orphans-only`	off	Skip CO findings, just list orphans
`--volume-ids`	none	Limit scope to these IDs
`--volume-ids-file`	none	Same, read from file
`--apply`	off	Call `ec2:ModifyVolume`
`--apply-all`	off	Required for `--apply` without an ID list
`--yes`	off	Skip interactive confirm prompt

Output formats

The script produces either a CSV (default) or a styled Excel workbook based on the file extension passed to --output.

Run manifest header

Every report (CSV and XLSX) carries a manifest banner at the top with full run provenance:

Generated 2026-05-21T18:23:00Z by ebs-cost-optimizer v1.8.0 |
Account 711457211352 | Region us-east-1
Parameters: lookback=30d, buffer=20.0%, floors=3000 IOPS / 125 MiB/s,
gp3-only=True, direction=both, applied=False
Pricing (gp3 USD): storage=$0.08/GiB-mo, iops=$0.005/IOP-mo over 3000,
throughput=$0.04/MiB/s-mo over 125

This makes a report self-describing: anyone receiving the file by email or in a ticket can see exactly which account, region, parameters, and pricing model produced it. In CSV, banner lines start with # so most tools ignore them by default. In XLSX, the banner sits in merged rows above the column headers.

Output filename and timestamps

By default, every run produces a uniquely named workbook in the form EBS_Cost_Optimization_<UTC date>_<UTC time>Z.<ext>:

EBS_Cost_Optimization.xlsx       ->  EBS_Cost_Optimization_2026-05-21_18-23-00Z.xlsx
EBS_Cost_Optimization.csv        ->  EBS_Cost_Optimization_2026-05-21_18-23-00Z.csv
reports/q2-audit.xlsx            ->  reports/q2-audit_2026-05-21_18-23-00Z.xlsx

The format is human-readable, lexically sortable, and Windows-safe (no :).

For explicit positioning, use the {ts} placeholder:

python3 ebs_cost_optimizer.py --region us-east-1 --output snapshot_{ts}.xlsx
# -> snapshot_2026-05-21_18-23-00Z.xlsx

To disable timestamping (overwrite same filename each run):

python3 ebs_cost_optimizer.py --region us-east-1 --no-timestamp \
    --output EBS_Cost_Optimization.csv

CSV

Plain text, suitable for downstream tooling, scripting, and version control.

python3 ebs_cost_optimizer.py --region us-east-1 \
    --output EBS_Cost_Optimization.csv

Styled Excel workbook (.xlsx, default)

Professional-grade workbook with frozen header, AutoFilter on every column, USD currency formatting (negative values in red), color-coded direction (green=downsize, yellow=upsize, peach=orphan, grey=no-change/skip), and visually distinct SUBTOTAL (pale blue) and GRAND_TOTAL (navy) rows.

python3 ebs_cost_optimizer.py --region us-east-1
# -> EBS_Cost_Optimization_2026-05-21_18-23-00Z.xlsx

Requires openpyxl (see Prerequisites above for install commands).

Output schema

Column	Description
`volume_id`	EBS volume ID, or `SUBTOTAL` / `ORPHAN_SUBTOTAL` / `GRAND_TOTAL`
`volume_type`	`gp3`, `io1`, etc.
`size_gib`	Volume size
`state`	`in-use`, `available`, etc.
`attached_instances`	Comma-separated EC2 instance IDs
`create_time`	ISO 8601
`current_iops` / `current_throughput_mibps`	Current provisioning
`peak_iops` / `peak_throughput_mibps`	Observed peak over `--days` window
`target_iops` / `target_throughput_mibps`	Recommended target
`direction`	`UPSIZE`, `DOWNSIZE`, or `NONE`
`action`	Human-readable summary
`monthly_cost_current_usd`	Estimated monthly cost at current settings
`monthly_cost_target_usd`	Estimated monthly cost at target
`monthly_delta_usd`	`target − current`. Negative = savings
`annual_delta_usd`	`monthly_delta × 12`
`modification_state`	Result of `ModifyVolume` (or `DRY_RUN`)
`co_finding_reasons`	Compute Optimizer reason codes
`category`	`rightsizing`, `orphan`, `subtotal`, or `total`
`notes`	Datapoint count, orphan age, encryption status

Safety controls

Read-only by default. --apply requires either --volume-ids / --volume-ids-file or explicit --apply-all.
Interactive confirmation before any modification. Non-interactive sessions require --yes.
Cooldown awareness. Skips volumes already mid-modification (the AWS 6-hour cooldown).
Adaptive retries for Compute Optimizer and CloudWatch throttling.
CSV injection guard for spreadsheet safety.
Atomic file writes so a crash never produces a half-written CSV.
Strict input validation for region, volume IDs, and numeric ranges.
Bounded parallelism to stay under CloudWatch account RPS limits.

Limitations

AWS Compute Optimizer takes ~24h after opt-in to populate findings, and ~14d to populate findingReasonCodes for new volumes.
CloudWatch retains 5-minute datapoints for 63 days; lookback is capped at 63.
The pricing model covers gp3 only. For io1/io2/gp2/st1/sc1, the cost columns remain at 0; widen the model in DEFAULT_PRICING_* if needed.
AWS only allows one ModifyVolume per volume every 6 hours. Re-runs within that window will skip cooldown-bound volumes.
Per-account by design. Multi-account scans require an STS AssumeRole wrapper.

Roadmap

Multi-account / Organizations-wide aggregator with AssumeRole loop.
Tag-based exclusion (do-not-modify=true).
AWS Lambda packaging with EventBridge schedule and SNS notifications.
Slack / Microsoft Teams notifier hook.
Optional Cost Explorer integration to pull actual blended rates per account.

Security

See SECURITY.md for the full security posture mapped to the AWS Well-Architected Security Pillar (identity, access, detection, network, compute, data protection, input validation, and apply-path safeguards).

For vulnerability reports, follow the AWS Vulnerability Reporting process described in CONTRIBUTING.md.

License

This project is licensed under the Apache-2.0 License. See LICENSE.

Disclaimer

This is sample code intended to demonstrate AWS service capabilities. Test in a non-production environment before applying changes to production volumes. The pricing defaults are illustrative; verify against your AWS billing rates before relying on the financial impact figures.

Name		Name	Last commit message	Last commit date
.gitignore		.gitignore
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
NOTICE		NOTICE
README.md		README.md
SECURITY.md		SECURITY.md
ebs_cost_optimizer.py		ebs_cost_optimizer.py
requirements.txt		requirements.txt
setup.sh		setup.sh

Navigation Menu

Repository files navigation

Amazon EBS Cost Optimizer

Overview

Why this is helpful

Use cases

Cost optimization (downsize)

Performance remediation (upsize)

Idle resource cleanup (orphans)

Multi-account governance

How it works

Data flow

Volume types

Prerequisites

Install the dependencies

Required IAM permissions

Quickstart

Common workflows

CLI reference

Output formats

Run manifest header

Output filename and timestamps

CSV

Styled Excel workbook (.xlsx, default)

Output schema

Safety controls

Limitations

Roadmap

Security

License

Disclaimer

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

Amazon EBS Cost Optimizer

Overview

Why this is helpful

Use cases

Cost optimization (downsize)

Performance remediation (upsize)

Idle resource cleanup (orphans)

Multi-account governance

How it works

Data flow

Volume types

Prerequisites

Install the dependencies

Required IAM permissions

Quickstart

Common workflows

CLI reference

Output formats

Run manifest header

Output filename and timestamps

CSV

Styled Excel workbook (.xlsx, default)

Output schema

Safety controls

Limitations

Roadmap

Security

License

Disclaimer

About

Resources

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages