netflix.ocelli.loadbalancer.RandomWeightedLoadBalancer Maven / Gradle / Ivy
package netflix.ocelli.loadbalancer;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.NoSuchElementException;
import java.util.Random;
import java.util.concurrent.atomic.AtomicReference;
import netflix.ocelli.LoadBalancer;
import netflix.ocelli.loadbalancer.weighting.ClientsAndWeights;
import netflix.ocelli.loadbalancer.weighting.WeightingStrategy;
import rx.Subscriber;
/**
* Select the next element using a random number.
*
* The weights are sorted such as that each cell in the array represents the
* sum of the previous weights plus its weight. This structure makes it
* possible to do a simple binary search using a random number from 0 to
* total weights.
*
* Runtime complexity is O(log N)
*
* @author elandau
*
*/
public class RandomWeightedLoadBalancer extends LoadBalancer {
public static RandomWeightedLoadBalancer create(final WeightingStrategy strategy) {
return new RandomWeightedLoadBalancer(strategy);
}
private final AtomicReference> clients;
public RandomWeightedLoadBalancer(final WeightingStrategy strategy) {
this(strategy, new AtomicReference>(new ArrayList()));
}
RandomWeightedLoadBalancer(final WeightingStrategy strategy, final AtomicReference> clients) {
super(new OnSubscribe() {
private final Random rand = new Random();
@Override
public void call(final Subscriber s) {
List local = clients.get();
final ClientsAndWeights caw = strategy.call(local);
if (!caw.isEmpty()) {
int total = caw.getTotalWeights();
if (total == 0) {
s.onNext(caw.getClient(rand.nextInt(caw.size())));
s.onCompleted();
}
else {
int pos = Collections.binarySearch(caw.getWeights(), rand.nextInt(total));
if (pos >= 0) {
pos = pos+1;
}
else {
pos = -(pos) - 1;
}
s.onNext(caw.getClient(pos));
s.onCompleted();
}
}
else {
s.onError(new NoSuchElementException("No servers available in the load balancer"));
}
}
});
this.clients = clients;
}
@Override
public void call(List t) {
this.clients.set(t);
}
}