Throttling

-o0O0o-

How can a system manage resource consumption to maintain performance and reliability?

-o0O0o-

Original author: Microsoft’s Azure Architecture Center

UML Model File:

Download Throttling Pattern

Model Format Visual Paradigm

Control the amount of resources by a specific appliacation instance. It allows the system to maitain its functionality and restrict some resources if necessary.

Advantages

Avoiding system overloading by limiting the demands during times guaranteeing availability and performance.
Guarantees that ressources are distributed fairly among features.
Mitigation of DoS risks by limiting the number of requests executed.
Gradual scaling by acting as a temporary buffer while ressources are scaled to meet demand.

Disadvantages

Reduce user experience as a result form throttling which could reject or postpone requests.
Complexity in implementation because it requires planning and reliable monitoring systems.

Code Examples

public class Throttler {
    private final long intervalInMillis;
    private long lastExecutionTime = 0;

    public Throttler(int maxCallsPerSecond) {
        if (maxCallsPerSecond <= 0) {
            throw new IllegalArgumentException(“maxCallsPerSecond must be greater than zero.”);
        }
        this.intervalInMillis = 1000L / maxCallsPerSecond;
    }

    public synchronized boolean allow() {
        long currentTime = System.currentTimeMillis();
        if (currentTime – lastExecutionTime >= intervalInMillis) {
            lastExecutionTime = currentTime;
            return true;
        }
        return false;
    }
}

public class Throttler {

private final long intervalInMillis;

private long lastExecutionTime = 0;

public Throttler(int maxCallsPerSecond) {

if (maxCallsPerSecond <= 0) {

throw new IllegalArgumentException(“maxCallsPerSecond must be greater than zero.”);

}

this.intervalInMillis = 1000L / maxCallsPerSecond;

}

public synchronized boolean allow() {

long currentTime = System.currentTimeMillis();

if (currentTime – lastExecutionTime >= intervalInMillis) {

lastExecutionTime = currentTime;

return true;

}

return false;

}