Spring-cloud-starter-ribbon provides client-side software load balancing algorithms

What is the ribbon?

The Ribbon is an open source project by Netflix that provides a client-side software load-balancing algorithm that connects Netflix's middle-tier services. The Ribbon client component provides a complete set of configuration items such as connection timeouts, retries, and so on. Simply put, all the machines behind load Balancer (lb) are listed in the configuration file, and the Ribbon automatically helps you to connect the machines based on certain rules (such as simple polling, then connecting, and so on). It is also easy to implement a custom load balancing algorithm using the Ribbon.

LB Program Classification

The current mainstream LB scheme can be divided into two categories: a centralized lb, which is the use of a separate lb facility between the consumer and provider of the service (either hardware, such as F5, or software, such as Nginx), which is responsible for forwarding the access request through some policy to the provider of the service, and the other in-process lb, Integrating the LB logic into the consumer, the consumer learns from the service registry which addresses are available and then chooses a suitable server from those addresses themselves. The ribbon belongs to the latter, which is just a class library, integrated in the consumer process, where the consumer obtains the address of the service provider.

Three: The main components and workflow of the Ribbon

The core groups of the Ribbon (all interface types) have the following:

ServerList

Used to get the address list. It can be either static (providing a fixed set of addresses) or dynamic (the list of addresses is queried periodically from the registry).

Serverlistfilter

Used only when using dynamic serverlist to use a certain policy in the original list of services to remove a subset of addresses.

IRule

Select a final service address as the LB result. The selection strategy has polling, weighting based on response time, breaker (when Hystrix is available), etc.

The ribbon prefers to use serverlist to get a list of all available services at work, then serverlistfilter a portion of the address, and finally selects a server as the final result in the remaining address through Irule.

IV: Introduction to the main load balancing strategy provided by the Ribbon

1. Simple poll Load balancer (roundrobin)

The request is polled in turn to schedule the different servers, that is, each time the schedule executes i = (i + 1) mod n, and the first server is selected.

2. Random Load Balancing

Randomly select server with a status of up

3. Weighted response Time Load balancing (weightedresponsetime)

Assign a weight according to the corresponding time, the longer the corresponding time, the smaller the weight, the lower the likelihood of being selected.

4. Zone-aware polling load balancing (zoneavoidancerule)

Compound to determine the performance of the server area and the availability of the server Select server

Ribbon self-load balancing strategy comparison

Policy Name	Policy statement	Policy description	Implementation Notes
Bestavailablerule	public class Bestavailablerule extends Clientconfigenabledroundrobinrule	Select a minimum of concurrent requests for the server	Examine the server individually, if the server is tripped, ignore it, and select the server activerequestscount the smallest
Availabilityfilteringrule	public class Availabilityfilteringrule extends Predicatebasedrule	Filter out back-end servers that are flagged as circuit tripped because they have failed to connect, and filter out those highly concurrent backend servers (active connections exceeds the configured threshold)	The logic of using a availabilitypredicate to contain the filtering server is to check the status of each server that is logged in the status
Weightedresponsetimerule	public class Weightedresponsetimerule extends Roundrobinrule	Assign a weight according to the corresponding time, the longer the corresponding time, the smaller the weight, the lower the likelihood of being selected.	A background thread periodically reads the evaluation response time from the status and computes a weight for each server. The calculation of weight is also relatively simple responsetime minus each server's own average responsetime is the weight of the server. Use Roubine policy to select server when the Statas is not formed when it is first run.
Retryrule	public class Retryrule extends Abstractloadbalancerrule	The retry mechanism on the selected load balancing policy machine.	During a configuration time period when the server selection is unsuccessful, try to select an available server using the Subrule method
Roundrobinrule	public class Roundrobinrule extends Abstractloadbalancerrule	Roundrobin Mode Polling Select server	Poll Index, select the server that corresponds to the index location
Randomrule	public class Randomrule extends Abstractloadbalancerrule	Randomly select a server	Randomly on index, select the server that corresponds to the index location
Zoneavoidancerule	public class Zoneavoidancerule extends Predicatebasedrule	Compound to determine the performance of the server area and the availability of the server Select server	Using Zoneavoidancepredicate and Availabilitypredicate to determine whether a server is selected, the previous decision determines whether the performance of a zone is available and rejects the unavailable zone (all servers). Availabilitypredicate is used to filter out servers with too many connections.

V. Use of

Introduction of POM

    < Dependency >        < groupId >org.springframework.cloud</groupId>        <  Artifactid>spring-cloud-starter-ribbon</artifactid>    </dependency>

Based on Resttemplate

Defined

 PackageCom.jsoft.testzookeeper.client.config;Importorg.springframework.cloud.client.loadbalancer.LoadBalanced;ImportOrg.springframework.context.annotation.Bean;Importorg.springframework.context.annotation.Configuration;Importorg.springframework.web.client.RestTemplate; @Configuration Public classConfig {@Bean @LoadBalanced  Publicresttemplate resttemplate () {return Newresttemplate (); }}

Note: Callout @loadbalanced to turn on load balancing, which is the default polling method.

Use

 PackageCom.jsoft.testzookeeper.client.service;ImportCom.netflix.hystrix.contrib.javanica.annotation.HystrixCommand;Importorg.springframework.beans.factory.annotation.Autowired;ImportOrg.springframework.stereotype.Service;Importorg.springframework.web.client.RestTemplate; @Service Public classHelloService { @Autowired Private resttemplate resttemplate; @HystrixCommand (Fallbackmethod= "Sayhellofallback")     Publicstring SayHello (string name) {returnResttemplate.getforentity ("http://service-zookeeper/hello?name=" + Name, String.class). GetBody (); }    Privatestring Sayhellofallback (string name) {return"Service Error"; }}

More configuration Reference Official documentation: Https://github.com/Netflix/ribbon/wiki/Features, such as the following rule configuration (not practiced):

springboot-h2. Load balancing strategy for RIBBON.NFLOADBALANCERRULECLASSNAME=COM.NETFLIX.LOADBALANCER.RANDOMRULE//Ribbon  

Description: Springboot-h2 for ClientName

Based on feign:

Feign clients do not have configuration items in the integration, mainly based on the configuration file to define the rules, examples of reference code, the following are mainly configuration file configuration.

In addition to the above configuration, you can specify the following:

Mainly based on the zookeeper dependency, the official reference: Http://cloud.spring.io/spring-cloud-zookeeper/multi/multi_ spring-cloud-zookeeper-dependencies.html, the sample configuration is as follows:

Spring.cloud.zookeeper.dependencies. service-zookeeper. required=truespring.cloud.zookeeper.dependencies. service-zookeeper. path=/service-zookeeperspring.cloud.zookeeper.dependencies. service-zookeeper. Loadbalancertype=round_robin

Description: Service-zookeeper is the specified alias and will eventually go to/service-zookeeper,loadbalancertype is the rule.

Summarize:

1, feel based on the ZK load balancing rules and the ribbon of the original rules two inseparable, the main looks as if the granularity is different, the global has been divided into different services and other settings.

2, the above conclusion is to be practiced. There will be a lot of things to be expected.

Maven Example:

Https://github.com/easonjim/spring-cloud-demo/tree/master/ZooKeeper

Reference:

Http://www.ccblog.cn/96.htm (The above internal part is transferred from this article)

Http://www.cnblogs.com/wangjing666/p/6985047.html

http://blog.csdn.net/w_x_z_/article/details/71156009 (load Balancing rule customization)

http://blog.csdn.net/shunhua19881987/article/details/75466797

http://www.idouba.net/netflix-source-ribbon-rule/

http://blog.csdn.net/daybreak1209/article/details/53582366

http://blog.csdn.net/liuchuanhong1/article/details/54728406

http://www.jianshu.com/p/7bb96e2dc944

Http://www.cnblogs.com/wangjing666/p/6994451.html

Spring-cloud-starter-ribbon provides client-side software load balancing algorithms