Understanding Load Balancing in Spring Cloud: Registering Multiple Services and Code Analysis

Load balancing (Load balancing) is a computer technology that distributes workload across multiple computers, network connections, CPUs, or other resources to optimize resource use, maximize throughput, minimize response time, and avoid overload.

Using a bank queue analogy, multiple service windows correspond to multiple service instances, and a ticket machine corresponds to a load balancer that assigns incoming requests to a specific instance.

In Spring Cloud, a single service may become a bottleneck under high concurrency, so registering multiple instances of the same service (e.g., setting server.port: 8091 while keeping the same spring.application.name) allows the load balancer to distribute traffic.

Example configuration for a second item service instance:

server:
  port: 8091  # service port
spring:
  application:
    name: hutao-microservice-item
eureka:
  client:
    registerWithEureka: true
    fetchRegistry: true
    serviceUrl:
      defaultZone: http://127.0.0.1:9090/eureka/
  instance:
    prefer-ip-address: true

A Feign client can be declared with @FeignClient(value = "hutao-microservice-item") to invoke the item service:

@FeignClient(value = "hutao-microservice-item")
@RequestMapping("/itemservice")
public interface FeignOrderService {
    @GetMapping("item/{itemId}")
    Items queryItem(@PathVariable("itemId") String itemId);
}

The DiscoveryClient interface provides methods such as List<ServiceInstance> getInstances(String serviceId) to retrieve all instances of a given service ID.

Spring Cloud’s load‑balancing mechanism works through LoadBalancerInterceptor and RibbonLoadBalancerClient. The interceptor intercepts RestTemplate calls, and the Ribbon client selects a server via loadBalancer.chooseServer, defaulting to a round‑robin rule.

To customize the algorithm, the IRule implementation can be changed in the configuration, for example:

hutao-microservice-item:
  ribbon:
    NFLoadBalancerRuleClassName: com.netflix.loadbalancer.RandomRule

After restarting, the service instances are chosen randomly instead of sequentially.

Key takeaways: register multiple identical service IDs on different ports or hosts; enable @LoadBalanced on RestTemplate to let Ribbon handle instance selection; use DiscoveryClient to query instances; and customize the load‑balancing rule via NFLoadBalancerRuleClassName.