INFO 2004 Chap.13 Scaling, Load Balancing and Monitoring
Scaling, Load Balancing and Monitoring
Unexpected demand becomes an architecture problem when a workload cannot continue to serve users predictably. A load balancer sits between users and servers and directs traffic rather than allowing the two sides to communicate directly. The course covers load-balancer types, differing capabilities, global server load balancing and scheduling algorithms, then asks how to choose.
Auto scaling changes capacity in response to conditions. Health checking matters because adding or retaining an unhealthy target does not improve service. Monitoring provides the metrics used to observe state and support scaling decisions; CloudWatch is named in the AWS delivery. The complete reasoning loop is therefore demand, observation, decision, capacity change, traffic distribution and renewed observation.
Keep vertical scale-up distinct from horizontal scale-out, healthy distinct from unhealthy and stateful distinct from stateless behaviour. These direction words are small but reverse the architecture when swapped.
What this chapter covers
- 01
Unexpected-demand scenarios
- 02
The load balancer as intermediary
- 03
Load-balancer types and capabilities
- 04
Scheduling algorithms
- 05
Global server load balancing
- 06
Auto scaling and health checks
- 07
Monitoring metrics and the feedback loop
Design a scaling feedback loop for a demand surge
- 1Observe demand and target health through monitoring metrics rather than guessing from user complaints alone.
- 1Remove or avoid unhealthy targets in the traffic-distribution decision.
- 1Use the scaling policy to change available capacity when its defined condition is met.
- 1Distribute traffic through the load balancer using a capability and algorithm suited to the workload.
- 1Continue monitoring after the change, because the response is a feedback loop rather than a one-off action.
Key terms
- Load balancer
- An intermediary that receives traffic and passes it to suitable servers or targets.
- Scheduling algorithm
- The rule used to choose which suitable target receives a request.
- Global server load balancing
- Traffic distribution across a broader geographic or site-level scope.
- Auto scaling
- Changing available capacity in response to defined conditions.
- Health check
- A test used to judge whether a target should receive or continue serving traffic.
- Metric
- An observed measure used to understand workload state and support a decision.
- CloudWatch
- The AWS monitoring service named in the delivered scaling and monitoring topic.
- Scale out
- Adding parallel capacity; keep it distinct from scale up, which increases the capability of existing capacity.
Scaling, Load Balancing and Monitoring FAQ
What role does a load balancer play?
It sits between users and servers and directs traffic to appropriate recipients.
Why do health checks matter?
They prevent traffic or capacity decisions from treating an unhealthy target as useful service capacity.
How are monitoring and auto scaling connected?
Monitoring supplies the observed state used by the scaling condition, and verifies the result after capacity changes.
Is a specific scaling threshold published here?
No. The chapter teaches the loop and routes implementation values to current Canvas resources.
What direction words should I check?
Scale up versus scale out, healthy versus unhealthy, stateful versus stateless, and inbound versus outbound.
Assessment move
Draw the feedback loop from memory, then explain every arrow in a sentence. Add a failure injection—an unhealthy target, misleading metric or unsuitable algorithm—and state where the loop should catch it. Practise qualitative design first; provider thresholds and quotas belong to the implementation evidence, not to remembered claims.
Working through Scaling, Load Balancing and Monitoring in INFO 2004? Sia is AskSia’s AI Computer Science tutor — ask any INFO 2004 Scaling, Load Balancing and Monitoring question and get a clear, step-by-step explanation grounded in how INFO 2004 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
