All cheat sheets
Databasescheat sheet

Redshift Reserved Nodes

Commit to 1 or 3 years of Redshift capacity for 40–75% off on-demand — one of the biggest discounts AWS offers on a service that's often a huge slice of the bill. Node-type- and region-locked, no refunds, compute-only — reserve the baseline.

Last reviewed: July 14, 2026

Prefer step by step? Take the guided Redshift Reserved Nodes walkthrough — free, anonymous, interactive.Have an AI assistant? Jump to the copy-paste audit prompt at the bottom of this sheet.

TL;DR: Redshift Reserved Nodes are one of the deepest discounts AWS offers — 40–75% off on a data warehouse that's often a giant slice of the monthly bill. You commit your cluster's node type for 1 or 3 years; the discount applies automatically to matching nodes. The constraints are strict: reservations are locked to node type and region, non-refundable, and compute-only. Capacity planning is the hard part, not the math — reserve the baseline, never the peak.

The numbers

  • Payment tiers: All Upfront up to ~75% off (3-yr), Partial Upfront ~60–65%, No Upfront ~40–50%.
  • Worked example — 4-node ra3.4xlarge in us-east-1, 24/7: on-demand $3.26/hr × 4 × 730 = $9,500/mo ($114K/yr); 1-yr All Upfront saves ~$39K in year one (~34%); 3-yr All Upfront saves ~$64K over three years (~56%). A 20-node cluster crosses six figures/yr.
  • Coverage is count-based: reserve 4, run 6 → first 4 reserved, other 2 on-demand; run 2 → still pay for 4 (no refunds).
  • Field example: a stable 12-node ra3.4xlarge production warehouse locked 3-yr All Upfront for ~56% off — $250K+ over three years; a brand-new evolving 6-node cluster correctly stayed on-demand.

Do this

  1. Confirm 90+ days of steady node-hours in Cost Explorer before committing — reserve against reality, not intentions.
  2. Reserve the baseline, not the peak — steady at 10 nodes, bursting to 15 → reserve 10, let 5 run on-demand at peak.
  3. Match node type and region exactly — an ra3.xlplus reservation does nothing for dc2.large or another region.
  4. Start 1-year if you're nervous — the deep 75% lives in 3-year terms, but a 1-year costs only ~10–15% of the savings and keeps options open as AWS ships new node families.
  5. Set a renewal reminder 60–90 days before expiry — reservations don't auto-renew and the bill can silently jump 2–3× overnight when they lapse.

Gotchas

  • Node type is set in stone — migrate ra3.xlplus → ra3.4xlarge (or to Serverless) mid-term and you pay the old reservation and full on-demand on the new nodes; only AZ-within-region is modifiable.
  • Compute only — backup storage, manual snapshots in S3, data-transfer-out, and Redshift Spectrum queries all still bill on-demand; even a 75% RI won't zero the bill.
  • Watch the node-family roadmap — AWS keeps releasing better price-performance families; time multi-year commitments around the stability of the one you're on.
  • No clean exit — the Marketplace is clunky and usually a loss.

Skip this if

  • The workload is experimental/POC, seasonal, or the architecture is evolving (evaluating Snowflake/BigQuery, frequent resizing) — stay on-demand until it settles, then reserve only the baseline.
  • Usage is variable rather than steady 24/7 — Redshift Serverless bills per query and can overlay the unpredictable part; if the cluster is merely idle on a predictable schedule, Pause/Resume may fit better. Confirm baseline node-hours with AWS Cost Explorer before committing.

Run this audit with your AI assistant

Paste this into Claude, ChatGPT, or any agent that can run the AWS CLI with read-only credentials. It audits your account for exactly the waste this sheet describes — and changes nothing.

You are auditing an AWS account's Amazon Redshift clusters for Reserved
Node savings. Use the AWS CLI with READ-ONLY credentials. Do not create,
modify, or delete anything — report findings and recommended (unapplied)
fixes only.

1. Inventory: aws redshift describe-clusters — capture node type (e.g.
   ra3.4xlarge), node count, and whether each cluster runs 24/7. aws
   redshift describe-reserved-nodes — existing reservations and end dates.
2. Baseline: pull 90 days of node-hours per (region, node type) from
   Cost Explorer. The sustained floor is the reservable count.
3. Fit: flag steady 24/7 production warehouses as RI candidates; flag
   experimental/POC/seasonal/evolving clusters for on-demand or Redshift
   Serverless.
4. Savings math: on-demand $/node-hr × count × 730 vs reserved (All
   Upfront up to ~75% on 3-yr, Partial ~60-65%, No Upfront ~40-50%).
   Reserve baseline only; note compute-only (backups, manual snapshots,
   data transfer, Spectrum still on-demand).

Report a table: cluster | region | node type | steady count | current RI |
recommended reserve | term/payment | est. $/mo saved | notes (node-type
lock, expiry). Change nothing.
Works with any assistant that can run shell commands.

Want the guided version?

The Redshift Reserved Nodes walkthrough covers this topic interactively — it asks about your setup, branches to what’s relevant, and quizzes you on the tricky parts. Free and anonymous.

Start the walkthrough

More Databases cheat sheets

Amazon RDS Proxy

Connection count, not CPU, is often what forces an RDS upsize. RDS Proxy pools thousands of client connections (especially from Lambda) onto ~10–30 real ones, so a small instance stays small — the proxy fee typically covers 10–25% of the avoided upsize.

Athena Partition Projection

Instead of asking S3 which partitions exist before every query, Partition Projection lets Athena calculate the paths from a pattern you declare — near-instant query startup on tables with thousands of partitions, zero LIST metadata calls.

Athena Query Result Reuse

Athena caches identical query results in S3, so a repeated dashboard refresh or scheduled report returns instantly with zero bytes scanned. A workgroup checkbox that routinely cuts thousands off a scan-heavy bill.

Aurora I/O-Optimized

Aurora bills a sneaky third bucket — I/O operations at ~$0.20/million — that can dwarf compute and storage on I/O-heavy workloads. I/O-Optimized zeroes it out for a ~30–40% higher instance rate. A break-even calc, not a default.