Load Balancing

• Client to gateway: DNS parse domain to different ip addresses. (If we have multiple data centers, DNS will find the geographically closest data center)

• Gateway to web server: reverse proxy

• web server to application server: connection pool

• database load balancing: partition / sharding

How to distribute traffic:

• random, round-robin

• min connection count

• min latency

• based on ip.

What if load balancer crashed?

TODO