Route 53 & DNS Fundamentals

The plain-language idea

Route 53 is AWS's DNS service: it answers "what does this domain name actually point to?" A few record types cover almost everything a frontend developer runs into:

• A record — maps a name directly to an IP address. The original, simplest DNS record.
• CNAME — maps a name to another name, which then gets resolved in turn. Useful, but DNS rules forbid a CNAME at a zone apex (the bare root domain, example.com with no subdomain), because the apex also has to hold other required records (like MX for mail) that can't coexist with a CNAME there.
• ALIAS — a Route 53-specific record type built to solve exactly that gap: it behaves like a CNAME (point this name at another AWS resource) but is legal at the zone apex, and resolves internally within Route 53 rather than requiring an extra DNS lookup. It only works for specific AWS resource types (a CloudFront distribution, a load balancer, an S3 website endpoint), not arbitrary hostnames.

On top of records, routing policies decide which answer to hand back when there's more than one candidate: simple (one static answer, no logic), weighted (split traffic across multiple answers by assigned proportion), latency-based (send each resolver toward whichever region is fastest for it), and failover (serve a primary answer, switch to a secondary the moment a health check reports the primary unhealthy).

See it in a real thing: pointing a bare domain at CloudFront

example.com — the apex, no www. — needs to point at a CloudFront distribution. A plain CNAME record can't do this; the DNS spec doesn't allow one at the apex. Route 53 solves it with an ALIAS record: example.com gets an ALIAS record pointing at the CloudFront distribution's domain name, which resolves correctly at the apex and, unlike a CNAME, doesn't add an extra hop at query time. www.example.com gets its own ALIAS (or CNAME) pointing at the same distribution, so both the bare domain and the www version work.

What people usually miss

Watch out

• ALIAS isn't a standard DNS record type — it's an AWS-specific construct, and it only works for a specific set of AWS resource types. It solves the "CNAME at the apex" problem, but reaching for it on an arbitrary non-AWS hostname doesn't work the same way.
• TTL controls how long a cached answer sticks around, and forgetting to lower it before a planned cutover is a real, recurring mistake. If a record's TTL is a day and it changes five minutes before a migration, resolvers that already cached the old answer keep handing it out for up to a day — the record changed correctly, but the effect of that change is delayed by whatever was already cached elsewhere.
• A failing health check doesn't remove a stale answer instantly, worldwide. Failover routing switches Route 53's own answer once a health check reports failure, but any resolver that already cached the old (healthy) answer keeps using it until that cached entry's TTL expires — the same caching lag that affects any DNS change.
• Weighted or latency-based routing isn't the same mechanism as a load balancer, even though the visible effect looks similar. DNS-level splitting happens once per (uncached) lookup, and a client that caches its DNS answer keeps hitting the same target for the life of that cache — a load balancer, by contrast, makes a fresh routing decision on every connection.

Why this matters in interviews

"How would you point a bare domain at a CloudFront distribution, when CNAME records can't be used at the zone apex?" is a real, specific question, and naming ALIAS by name — not just "some kind of DNS trick" — is what separates someone who's actually configured DNS on AWS from someone who's only used it as a black box.