Balancing across the planet
A local balancer spreads load inside one data center. Global server load balancing, or GSLB, decides which data center a user reaches in the first place, steering traffic across regions for latency, capacity, and disaster recovery.
How GSLB steers
GSLB commonly works through DNS or anycast.
- DNS based GSLB: the authoritative resolver returns different backend addresses depending on the client's location, the health of each site, and current load.
- Anycast based GSLB: the same address is announced from many sites, and routing carries the user to the nearest one.
Inputs to the decision
- Geo proximity: pick the closest site to cut latency.
- Health: drop sites that fail their checks.
- Capacity and load: shift traffic away from a saturated region.
- Policy: honor data residency or failover order.
The DNS caveat
DNS based GSLB inherits DNS limits. Resolver caching and TTL mean changes take time to propagate, and the resolver's location may differ from the user's, so geo decisions can be coarse. Short TTLs improve agility but raise query volume and reduce cache benefit.
Key idea
GSLB chooses which data center a user reaches using geo proximity, health, and load, typically via DNS or anycast, with DNS based designs limited by caching and TTL.