DNS Fundamentals

How a domain name becomes an IP address and why DNS matters at every layer of scale.

IntermediateTheory10 min read

Read these first

Zero to Millions of Users

How DNS resolution works

When a browser requests api.streamhub.com, it does not know the server's IP. A multi-step lookup resolves the name.

DNS lookup path (Route 53)

Every request starts here — api.streamhub.com resolves to ALB or CloudFront.

StreamHub production architecture (AWS)

End-to-end path from user to data — reference this when placing any new service.

Resolution chain

Browser cache — recent lookups cached for TTL seconds.
OS resolver cache — local stub resolver.
Recursive resolver — ISP DNS or public resolver (8.8.8.8, 1.1.1.1).
Root → TLD → authoritative — delegation until the domain's nameserver answers.
A/AAAA record — IPv4 or IPv6 address returned to the client.

Java

# Trace a lookup (simplified)
dig +trace api.streamhub.com A

# Check TTL — how long clients cache this answer
dig api.streamhub.com | grep -E 'ANSWER|IN A'

DNS record types you need

Record	Purpose	Example
A / AAAA	Hostname → IP	api.streamhub.com → 52.12.34.56
CNAME	Alias to another name	www → streamhub.com
MX	Mail server	mail.streamhub.com
TXT	Verification, SPF, DKIM	v=spf1 include:...
NS	Authoritative nameserver	ns1.cloudprovider.com
SRV	Service location	_https._tcp service discovery

A / AAAA
PurposeHostname → IP
Exampleapi.streamhub.com → 52.12.34.56
CNAME
PurposeAlias to another name
Examplewww → streamhub.com
MX
PurposeMail server
Examplemail.streamhub.com
TXT
PurposeVerification, SPF, DKIM
Examplev=spf1 include:...
NS
PurposeAuthoritative nameserver
Examplens1.cloudprovider.com
SRV
PurposeService location
Example_https._tcp service discovery

For system design interviews, A/AAAA, CNAME, and TTL matter most.

TTL and caching trade-offs

TTL (time to live) controls how long resolvers cache a record.

Aspect	Low TTL (60s)	High TTL (3600s)
Failover speed	Fast — clients pick up new IP quickly	Slow — stale IPs during migration
DNS load	More queries to authoritative NS	Fewer queries, lower cost
Use case	Blue-green deploys, DR drills	Stable CDN CNAMEs

Failover speed
Low TTL (60s)Fast — clients pick up new IP quickly
High TTL (3600s)Slow — stale IPs during migration
DNS load
Low TTL (60s)More queries to authoritative NS
High TTL (3600s)Fewer queries, lower cost
Use case
Low TTL (60s)Blue-green deploys, DR drills
High TTL (3600s)Stable CDN CNAMEs

StreamHub uses 60-second TTL on API endpoints during migrations and 300 seconds in steady state.

DNS in scaled architectures

DNS roles at scale

Load balancing — multiple A records (round-robin DNS) or CNAME to an LB hostname.
Geographic routing — Route 53 latency-based or geolocation policies send users to the nearest region.
CDN — CNAME cdn.streamhub.com → d111111abcdef8.cloudfront.net.
Service discovery — internal DNS (CoreDNS in Kubernetes) maps streamhub-api.default.svc.cluster.local.

Java

# Example Route 53 weighted routing during canary
api.streamhub.com:
  - type: A
    alias: lb-us-east-1
    weight: 90
  - type: A
    alias: lb-us-east-1-canary
    weight: 10

DNS as a failure point

DNS outages take down services that are otherwise healthy. Mitigations:

Use managed DNS with anycast (multiple global PoPs).
Keep TTLs low enough for failover but not so low that you DDoS yourself.
Monitor resolution time and NXDOMAIN rates.
Have a runbook for nameserver migration (pre-lower TTL 24h before change).

Java

# Health check from multiple regions
curl -w '%{time_namelookup}\n' -o /dev/null -s https://api.streamhub.com/health

Internal vs external DNS

External DNS faces the public internet. Internal DNS resolves service names inside a VPC or Kubernetes cluster — critical for microservice-to-microservice calls without hard-coded IPs.

Java

External:  api.streamhub.com     → public ALB
Internal:  video-service.prod     → 10.0.4.22 (pod IP, short TTL)

Quick recall

Everything you need if you only revisit this box.

DNS maps hostnames to IPs through recursive and authoritative resolvers.
A/AAAA for IPs, CNAME for aliases; TTL balances failover speed vs query load.
Low TTL for migrations; higher TTL for stable CDN endpoints.
DNS enables geo-routing, CDN CNAMEs, and internal service discovery.
DNS failure makes healthy servers unreachable — use managed anycast DNS.
Always include DNS in end-to-end traffic diagrams.

Test yourself

Answer these before moving on — recall is what makes it stick.