Skip to content

Commitments: Savings Plans, Reserved Instances, Azure Reservations and GCP CUDs

The Commitments page answers four questions about the discounts you have already bought: what you own, how well it is used, when it ends, and what to buy next.

It covers AWS Savings Plans and Reserved Instances, Azure reservations and Azure savings plans, and GCP committed use discounts (CUDs).

Coverage and utilization are different things

Section titled “Coverage and utilization are different things”

These two numbers get confused constantly, and they fail in opposite directions.

Metric Question it answers What a bad number means
Coverage How much of my eligible usage is running at a discounted rate? Low coverage means you are paying on demand for steady workloads. Buy more.
Utilization How much of what I committed to am I actually using? Low utilization means you are paying for a commitment you are not consuming. You bought too much, or the workload moved.

You want both high. High coverage with low utilization means you over committed. Low coverage with high utilization means you under committed and have room to buy more.

Utilization is cost weighted, not an average of percentages. A 100 unit commitment at 50 percent and a 1 unit commitment at 100 percent scores 50.5 percent, not 75 percent, because the big commitment dominates the money at stake.

The page targets 80 percent utilization, shown as a dashed line on the chart and as the colour threshold in the tables.

Pick a range of 30, 60 or 90 days.

Stat cards

  • Coverage, the share of eligible usage covered over the range
  • Utilization, against the 80 percent target
  • Effective savings rate, the share of your on demand bill saved
  • Unused commitment, money committed but not consumed
  • Expiring soon, counts in the next 30 and 90 days

Utilization and coverage trend plots daily utilization and coverage over the range, with the 80 percent reference line. Tabs switch between All and each family (see below), and a table lists each family and service with its commitment count, utilization, coverage and unused amount. Click a family to chart it on its own.

Inventory has tabs for Active, Expiring in 90 days and All, with columns for commitment, term, scope, monthly cost, utilization over 30 days, and end date. Expand a row to see that commitment’s own history (below).

Expiry calendar shows the next 12 months, one tile per month (see Expiry calendar).

Purchase recommendations has tabs for All terms, 1 year and 3 years, with columns for option, term and payment, upfront, monthly commitment, monthly savings and break even.

Options for the same usage are alternatives. Pick one per row group rather than adding them up, or you will double count the savings.

Break even shows Immediate when an option has no upfront payment, because there is nothing to pay back.

Xplorr stores one utilization figure per commitment per day, and one coverage figure per family per day, so you can see a trend instead of a single number for the range.

Families. Each provider’s commitments are grouped into two families, which is also how the providers split eligible usage for coverage:

Family Includes
Savings plans AWS Savings Plans, Azure savings plans, GCP spend based CUDs
Reservations AWS Reserved Instances, Azure reservations, GCP resource based CUDs

Within a family, utilization is also broken down by service, for example an RDS Reserved Instance.

How coverage is measured. AWS reservation coverage is in hours: reserved hours out of eligible hours, as Cost Explorer reports it. Every other family’s coverage is in spend: the committed share of eligible spend. The providers’ daily coverage data is not split by service, so services show utilization only.

Per commitment history. Expanding an inventory row shows, for the page’s range:

  • Covers, what the commitment applies to
  • Utilization over the range
  • Unused, the amount committed but not used
  • Net savings
  • Days with data, for example 28 of 30
  • A daily chart of used and unused amounts with the utilization line and the 80 percent target

Days that were estimated rather than reported (see below) are marked as such.

The Expiry calendar covers the next 12 months, one tile per month. Each tile shows how many commitments end that month and the monthly commitment cost that ends with them (or Mixed currencies). Each commitment shows its end date, what it covers, its monthly cost and its account.

Commitments ending within 30 days are highlighted red with the number of days left (or Today), and those within 90 days amber. The card’s heading counts the total, and how many end within 30 and within 90 days.

Not every capability exists for every provider, and the gaps matter when you plan.

Capability AWS Azure GCP
Inventory Yes Yes Yes
Coverage Yes Yes Yes
Daily utilization Yes, from Cost Explorer Yes, from amortized cost; estimated where that is not readable Estimated
Purchase recommendations Yes Yes Yes

What each provider gives:

  • AWS: Savings Plans and Reserved Instance utilization per day from Cost Explorer. Cost Explorer only reports Savings Plans utilization for days before the current date, so the newest day stored is two days ago. A day Cost Explorer has no data for is stored as no data rather than failing the rest.
  • Azure: reservation and savings plan utilization per day from amortized cost on the subscription: used benefit against UnusedReservation or UnusedSavingsPlan charges, in money. This needs only Cost Management Reader on the subscription, so utilization is stored even when the tenant scope listings are not readable. A commitment found this way but not in the listings is named from its ID, without a term or end date, until the listing becomes readable. Where amortized cost is not readable, Xplorr falls back to an estimate from reported reservation hours and the commitment price.
  • GCP: estimated. GCP reports committed use credits per region, so Xplorr derives cost from the commitment price.

The page marks estimated figures and shows a disclaimer when any are in view.

Azure purchase recommendations cover reservations (shared and single subscription scope) and savings plans (1 and 3 year terms), from the Azure Consumption and Cost Management recommendation APIs over a 30 day lookback. They need Reader or Cost Management Reader on the subscription. GCP purchase recommendations cover resource based CUDs per region and, when a billing account ID is set on the account, spend based CUDs; they need Recommender Viewer on the project, the Recommender API enabled, and Billing Account Viewer on the billing account for spend based CUDs. If a permission is missing, the Commitments page and the Data health page name the exact permission and role to add; the rest of the sync keeps working.

Commitment data needs a few permissions beyond the basic cost connection. The page lists them per provider when it has no data yet.

AWS

ce:GetSavingsPlans*
ce:GetReservation*
savingsplans:DescribeSavingsPlans
ec2:DescribeReservedInstances
rds:DescribeReservedDBInstances
elasticache:DescribeReservedCacheNodes

Azure:

  • Cost Management Reader (or Reader) on the subscription, which the cost connection already needs, covers coverage, daily reservation and savings plan utilization (from amortized cost), and purchase recommendations.
  • Reservations Reader and Savings plan reader, both assigned at tenant scope, add each commitment’s term, price and end date. Owned reservations and savings plans are tenant level resources, so no subscription role lists them. Assigning a role at tenant scope needs User Access Administrator there; see Connect Azure. Without these, the expiry calendar cannot show an end date for those commitments.

A refused permission is recorded and shown; it never fails the sync.

GCP: Compute Viewer, Recommender Viewer, and read access to the billing export.

The commitment sync runs daily at 03:30 UTC, after the cost sync.

  • The first run backfills 30 days of history.
  • Later runs refetch the last 3 days, because providers keep revising recent days. After a missed run, a run fetches enough days to close the gap, up to 30.
  • Coverage and utilization are planned separately. An account that holds a running commitment but has no stored utilization yet is backfilled for 30 days.
  • Purchase recommendations refresh on a 7 day cycle.

Admins can click Sync now. Only one commitment sync runs at a time, so if another is already in progress you will be told to wait rather than starting a second.

When Azure utilization comes from the estimate fallback rather than amortized cost, savings plan history builds up one day per run instead of backfilling.

Commitments are graded by how close they are to ending:

Window Status
Already past the end date Expired
7 days or fewer Critical
8 to 30 days Warning
31 to 90 days Notice
More than 90 days No badge

A commitment with no end date on record shows No end date.

An expiry is not a small event. When a reservation lapses, that usage reverts to on demand pricing the same day, which shows up as a step change in your bill rather than a gradual drift.

Two commitment alerts are raised automatically and appear with your other alerts.

Expiring: fires 30 days out and again 7 days out. The 7 day alert is high severity.

Low utilization: fires when a commitment averages below 80 percent over the last 7 days, and needs at least 3 days of data before it will fire at all. Below 50 percent it is high severity.

Each alert is raised once per commitment per month, so resolving one does not bring it back the next day, and an unresolved expiry re-raises the following month.

“No commitment data yet”

The sync has not run since you connected the account, or the permissions above are missing. The sync runs daily at 03:30 UTC. Admins can force it with Sync now.

Utilization looks wrong for Azure or GCP

GCP utilization, and Azure utilization when amortized cost is not readable, is estimated. Compare the trend rather than the exact figure. For Azure, check that the service principal has Cost Management Reader on the subscription.

The expiry calendar is missing an Azure commitment, or it has no end date

The end date comes from the tenant scope listings. Grant Reservations Reader and Savings plan reader at tenant scope.

A commitment is missing from the inventory

Some AWS reservation types are reconstructed from Cost Explorer rather than read from a direct API. If a commitment exists in your provider console but not here, check that the matching Describe permission is granted.

Recommendations disappeared after a sync

They do not. Stale recommendations are only cleared after a complete refresh, so a partial provider failure leaves the previous set in place.

Coverage is high but the bill did not fall

Check utilization. Coverage counts usage running under a commitment; it says nothing about whether you are using everything you committed to.

Can I buy a commitment from Xplorr? No. Xplorr is read only. It tells you what to buy and what it would save; you make the purchase in your provider console.

Why do recommendations show several options for the same workload? They are alternatives at different terms and payment options. Choose one per group.

Does the 80 percent target apply to everyone? It is the default target on the page. Steady production workloads can justify higher; bursty ones often should not be committed at all.

How does this differ from the Coverage tab? The Coverage tab on this page focuses on the on demand against committed split of your compute spend. The Commitments tab focuses on the commitment contracts themselves, their utilization and their expiry.