Skip to content

Notes on AWS, the parts nobody documents.

By Wellar Team.

Hand-drawn diagram: on the left, RDS Multi-AZ with a primary and a standby that can't serve reads, each on its own volume; on the right, Aurora with a writer and a reader sharing one volume copied six times across three Availability Zones.
20 min read

Aurora vs RDS: when Aurora is worth paying for

Aurora vs RDS for PostgreSQL and MySQL, priced on one 500 GB database: what moving to Aurora buys, when RDS stays cheaper, and how to migrate.

aws-costaurorards
Line chart of one S3 log bucket growing 1,000 GB a month for 36 months at us-east-1 prices: with no lifecycle policy the monthly bill climbs in a straight line to $828; with one rule it levels off at $115 from month 24.

S3 lifecycle policy: what having none costs, and the rules that lose money

What an S3 lifecycle policy does, what having none costs on cold data, a Terraform example, and the per-object fees that make a bad rule lose money.

19 min read
Editorial miniature showing a row of healthy backup modules connected by one amber test bridge to a running application module.

AWS Backup Restore Testing: A Backup Is Not a Recovery Plan

Use AWS Backup restore testing to schedule real recoveries, validate restored RDS and EBS resources, measure RTO, and catch failures before an incident.

12 min read
Editorial illustration of one cyan data stream crossing between two dark infrastructure zones, with an amber meter on each side of the boundary.

AWS Cross-AZ Data Transfer: Why One GB Shows Up Twice

Trace AWS cross-AZ data transfer cost from the billing line to the ENI path, explain the two-sided charge, and separate resilience from avoidable waste.

11 min read
Editorial diagram of a thick stream of log events entering a small storage reservoir, titled CloudWatch Logs cost, retention is not the first fix.

CloudWatch Logs Cost: Retention Is the Wrong First Fix

A practical CloudWatch Logs cost audit: separate ingestion from storage, find the noisiest log groups, and cut spend without deleting useful evidence.

12 min read
A dark home-office desk at night lit by laptop glow and a small lamp, with a phone on the desk showing a notification at 22:14 on a Sunday.

AWS bill too high? The 8 questions to ask before touching the console.

A senior engineer's triage order for an AWS bill that just spiked. Eight questions, ranked by dollars-per-minute. Most teams should stop at question five.

19 min read
Calculator-style graphic comparing a 500 GiB gp2 EBS volume at $50 per month with gp3 baseline at $40 and gp3 matched throughput at $45.

AWS gp2 to gp3: the 20% savings claim has a footnote

An AWS EBS gp2 vs gp3 migration guide with cost math, CloudWatch checks, and the gotcha that shrinks a 20% win to 8-10%.

11 min read
Side-by-side mock of an AWS Trusted Advisor check showing zero idle NAT Gateways alongside a single NAT Gateway tile labeled $48.32 over 41 days

Trusted Advisor is wrong about your NAT Gateways. Here's the math.

AWS's new Trusted Advisor check still misses a category of idle NAT Gateways that costs $32+/month each. Here's the math, the metric, and the CLI.

11 min read

You've reached the end. Thanks for reading.