Cloud Traffic Control
Survive the overload!
Your mission
It's exam day and the whole class logs in at once. Keep the online exam system alive: make a server bigger, add more servers, switch on a load balancer and automatic scaling, then survive a crashed server and a power cut in a whole data centre zone. Every server costs money, so don't just buy everything.
Explore
Turn up the number of students online and keep the exam system fast: make the server bigger, add servers, and let a load balancer share the work.
This is the university's online exam system. Every student sends requests to it, and a server sends back responses. Push the Students online slider up and watch what happens to one small server. Then fix it.
0 of 6 discoveries made
Experiment
Run the exam day from 13:30 to 16:30. Survive the rush, a crashed server and a power cut in a whole zone, and keep the bill down.
Exam-day missions
Try each one. They tick themselves off.
0 of 5 experiments done
Challenge
Six rounds: name the technique, plan the servers, and keep a service available when things fail.
Traffic controller's notes
- Vertical scaling: a bigger server
- Horizontal scaling: more servers
- Elasticity: up and down automatically with demand
- Load balancer + health checks: work only for healthy servers
- Redundancy, replication, failover → high availability
- Avoid a single point of failure
Finish all 6 rounds to finish this stage.
Understand
Here is what exam day was showing you, in the words you need for the exam.
Scalability and elasticity
Cloud systems can change their computing resources to meet different levels of demand.
| Technique | What it means | On exam day |
|---|---|---|
| Vertical scaling | Increase the resources of an existing server (more CPU, RAM or storage) | Server 1 grew from 2 CPU / 8 GB RAM to 8 CPU / 32 GB RAM |
| Horizontal scaling | Add more servers to handle a higher workload | From 1 server to 3 servers |
| Elasticity | Automatically adjust resources up or down as demand changes | More servers during the exam, fewer after it |
The load balancer
A load balancer distributes incoming requests among multiple servers. Without it, every student reached Server 1 and the extra servers sat idle. Multiple servers let the cloud handle many users at the same time, share the workload, and improve performance and availability.
Health checks
The load balancer regularly checks whether each server is healthy. If a server is unavailable, it sends new requests only to the healthy servers. With health checks switched off, it kept sending students to the crashed server.
When a whole zone fails
| Term | Meaning |
|---|---|
| Redundancy | Extra (duplicate) resources, so the service can continue if one fails |
| Replication | Copies of data kept on multiple servers or in different locations |
| Failover | Automatically switching users to a working server when a failure occurs |
| High availability | Keeping the service accessible for as much time as possible (e.g. 99.9% or higher) |
A spare server in Zone B was not enough on its own: it also needed a copy of the data (replication), and requests had to be switched to it (health checks and failover). A well-designed cloud system avoids depending on a single point of failure.
Why pay attention to cost?
Cloud resources are rented and paid for based on usage. Keeping the peak number of servers running all day costs more; elasticity adds servers when they are needed and removes them afterwards. Cloud costs depend on usage: flexible, but not always cheaper.
Exam-day numbers
In the game a small server (2 CPU, 8 GB RAM) handled about 100 requests per second, a medium one 200 and a large one 400. These numbers are illustrative; the ideas are what count.
Words to know
- Vertical scaling
- Making an existing server bigger.
- Horizontal scaling
- Adding more servers.
- Elasticity
- Adjusting resources up or down automatically as demand changes.
- Load balancer
- Distributes incoming requests among multiple servers.
- Health check
- A regular test of whether a server is working.
- Failover
- Automatically switching users to a working server.
- Replication
- Keeping copies of data in several places.
- Single point of failure
- One part whose failure stops the whole service.
Exam tip
Name each technique precisely: a bigger server is vertical scaling, more servers is horizontal scaling, and adding and removing resources automatically with demand is elasticity. For a failure question, explain the job of each protection separately: redundancy gives a spare server, replication gives it the data, health checks detect the failure, and the load balancer or failover sends new traffic to the healthy server.