​Approach

​Mechanism

Application scenario

​Limitations

​Cost-efficiency

​Horizontal scaling

​Adding new instances

​Sustained increase in load

Startup latency, overshoot, inter-node overhead​

​Moderate for short spikes, higher for sustained ones

​Load shedding

​Dropping part of the incoming flow

​Critical backlog growth, time constraints

​Partial data loss

​High for short critical spikes

​Admission control / throttling

Limiting incoming flow

​Early-stage load increase, risk of overload

Depends on accurate threshold definition, restricts part of requests

High for moderate and significant spikes

​Prediction-based / predictive scaling

​Proactive resource adjustment

​Predictable and recurring spikes

​Depends on forecast accuracy

​High for regular patterns

​Vertical scaling

​Increasing resources of existing instances

​Short sudden spikes

​Limited by node capacity

​High for short-term spikes