Skip to content

Domain Name System (DNS)

What it does

DNS translates a domain name (e.g., www.example.com) into an IP address.

How it works

  • DNS is hierarchical, with a few authoritative servers at the top level.
  • Your router or ISP tells you which DNS server(s) to contact for lookups.

  • Lower-level DNS servers cache mappings, which can become stale due to propagation delays.

  • Results can also be cached by your browser/OS for a period set by the time-to-live (TTL).

Common record types

Record Meaning
NS (name server) Specifies the DNS servers for your domain/subdomain.
MX (mail exchange) Specifies the mail servers for accepting messages.
A (address) Points a name to an IP address.
CNAME (canonical) Points a name to another name (or CNAME, or an A record).

Managed DNS and routing methods

Managed DNS services include Cloudflare and AWS Route 53. They can route traffic using:

  • Weighted round robin — avoid servers under maintenance, balance between cluster sizes, A/B testing.

  • Latency-based — route to the nearest/fastest region.

  • Geolocation-based — route based on user location.

Disadvantages of DNS

  • A DNS lookup adds a slight delay (mitigated by caching).

  • DNS server management is complex and generally managed by governments, ISPs, and large companies.

  • DNS is a target for DDoS attacks (e.g., the 2016 Dyn attack took down sites like Twitter for users who didn't know the IP directly).

Key takeaways

  • DNS is effectively a giant, hierarchical, cached key-value store mapping names to addresses.

  • TTL governs freshness vs. caching benefit — a longer TTL reduces lookups but slows propagation of changes.

  • It's an example of eventual consistency in the wild.

Common Interview Questions

Design a CDN like Cloudflare (global routing)

Why it's asked here: a CDN's global routing is fundamentally a DNS routing problem.

Key points to discuss:

  • Use geo/latency-based DNS to route users to the nearest PoP.
  • A/AAAA, CNAME, and NS records; the TTL trade-off (fast propagation vs caching).
  • Anycast to route one IP to the nearest node.
  • How DNS caching layers (browser, OS, resolver) affect staleness.
flowchart LR
    U[User] --> R[Resolver]
    R --> A[Authoritative DNS]
    A -->|geo / latency| E1[Edge POP 1]
    A -->|geo / latency| E2[Edge POP 2]
    U --> E1
Design a globally available multi-region service

Why it's asked here: routing traffic across regions relies on DNS.

Key points to discuss:

  • Weighted round-robin, latency-based, and failover routing policies.
  • Health checks to stop sending traffic to unhealthy regions.
  • DNS TTL and why failover can be slow (clients cache old IPs).
  • DNS as a control plane vs data plane trade-off.
flowchart LR
    U[User] --> D[DNS]
    D -->|weighted / latency| R1[Region A]
    D -->|health check| R2[Region B]
    R1 <-->|replication| R2

Further reading