Load Balancer vs Reverse Proxy
Hop's server has been doing everything itself: handling HTTPS, compressing responses, answering every browser directly. Soon there will be five servers. Both changes point to the same idea — put something in front — and two names get used for it, often loosely. Here is the difference.
The rule
A reverse proxy stands in front; a load balancer also chooses who serves.
A reverse proxy is a server that receives requests from clients and forwards them to back-end servers, then passes the answers back. Clients only ever see the proxy. That position is useful even with one back end:
- TLS termination — handle HTTPS certificates and encryption in one place.
- Compression and caching — shrink responses and answer repeat requests without bothering the back end.
- Hiding and protecting — the back end's address stays private; limits and filters sit at the door.
- One front, many back ends — send
/api/to one service and/to another.
A load balancer spreads requests across several back-end servers that do the same job. It picks a server for each request, stops sending to servers that fail health checks, and lets you add or remove servers without clients noticing.
So a load balancer is a reverse proxy with one main job: distributing. Many products (for example NGINX, HAProxy, Envoy and cloud load balancers) can do both.
browsers ──▶ [front door] ──▶ Hop server 1
├──▶ Hop server 2
└──▶ Hop server 3The shortcut
Count the servers behind the door. One → you need a reverse proxy, if anything. Two or more doing the same job → you need load balancing.
One server
A reverse proxy handles HTTPS and compression, and hides the server's address. Nothing to balance yet.
The trap is fixing the servers' single point of failure by creating a new one at the door.
Drills
Answer fast — trust the rule.
Hop has exactly one application server. Which of these still makes sense?
What does a load balancer do that a plain reverse proxy may not?
What happens when one of the servers behind a load balancer fails its health checks?
After adding a load balancer, what is the new single point of failure if there is only one?
Remember
- A reverse proxy stands in front; a load balancer also chooses who serves.
- Shortcut: one server → reverse proxy; two or more doing the same job → load balancing.
- Trap: a single front door becomes the new single point of failure.