Java實現(xiàn)平滑加權輪詢算法之降權和提權詳解
前言
上一篇講了普通輪詢、加權輪詢的兩種實現(xiàn)方式,重點講了平滑加權輪詢算法,并在文末留下了懸念:節(jié)點出現(xiàn)分配失敗時降低有效權重值;成功時提高有效權重值(但不能大于weight值)。
本文在平滑加權輪詢算法的基礎上講,還沒弄懂的可以看上一篇文章。
現(xiàn)在來模擬實現(xiàn):平滑加權輪詢算法的降權和提權
1.兩個關鍵點
節(jié)點宕機時,降低有效權重值;
節(jié)點正常時,提高有效權重值(但不能大于weight值);
注意:降低或提高權重都是針對有效權重。
2.代碼實現(xiàn)
2.1.服務節(jié)點類
package com.yty.loadbalancingalgorithm.wrr; /** * String ip:負載IP * final Integer weight:權重,保存配置的權重 * Integer effectiveWeight:有效權重,輪詢的過程權重可能變化 * Integer currentWeight:當前權重,比對該值大小獲取節(jié)點 * 第一次加權輪詢時:currentWeight = weight = effectiveWeight * 后面每次加權輪詢時:currentWeight 的值都會不斷變化,其他權重不變 * Boolean isAvailable:是否存活 */ public class ServerNode implements Comparable<ServerNode>{ private String ip; private final Integer weight; private Integer effectiveWeight; private Integer currentWeight; private Boolean isAvailable; public ServerNode(String ip, Integer weight){ this(ip,weight,true); } public ServerNode(String ip, Integer weight,Boolean isAvailable){ this.ip = ip; this.weight = weight; this.effectiveWeight = weight; this.currentWeight = weight; this.isAvailable = isAvailable; } public String getIp() { return ip; } public void setIp(String ip) { this.ip = ip; } public Integer getWeight() { return weight; } public Integer getEffectiveWeight() { return effectiveWeight; } public void setEffectiveWeight(Integer effectiveWeight) { this.effectiveWeight = effectiveWeight; } public Integer getCurrentWeight() { return currentWeight; } public void setCurrentWeight(Integer currentWeight) { this.currentWeight = currentWeight; } public Boolean isAvailable() { return isAvailable; } public void setIsAvailable(Boolean isAvailable){ this.isAvailable = isAvailable; } // 每成功一次,恢復有效權重1,不超過配置的起始權重 public void onInvokeSuccess(){ if(effectiveWeight < weight) effectiveWeight++; } // 每失敗一次,有效權重減少1,無底線的減少 public void onInvokeFault(){ effectiveWeight--; } @Override public int compareTo(ServerNode node) { return currentWeight > node.currentWeight ? 1 : (currentWeight.equals(node.currentWeight) ? 0 : -1); } @Override public String toString() { return "{ip='" + ip + "', weight=" + weight + ", effectiveWeight=" + effectiveWeight + ", currentWeight=" + currentWeight + ", isAvailable=" + isAvailable + "}"; } }
2.2.平滑輪詢算法降權和提權
package com.yty.loadbalancingalgorithm.wrr; import java.util.ArrayList; import java.util.List; /** * 加權輪詢算法:加入存活狀態(tài),降權使宕機權重降低,從而不會被選中 */ public class WeightedRoundRobinAvailable { private static List<ServerNode> serverNodes = new ArrayList<>(); // 準備模擬數(shù)據(jù) static { serverNodes.add(new ServerNode("192.168.1.101",1));// 默認為true serverNodes.add(new ServerNode("192.168.1.102",3,false)); serverNodes.add(new ServerNode("192.168.1.103",2)); } /** * 按照當前權重(currentWeight)最大值獲取IP * @return ServerNode */ public ServerNode selectNode(){ if (serverNodes.size() <= 0) return null; if (serverNodes.size() == 1) return (serverNodes.get(0).isAvailable()) ? serverNodes.get(0) : null; // 權重之和 Integer totalWeight = 0; ServerNode nodeOfMaxWeight = null; // 保存輪詢選中的節(jié)點信息 synchronized (serverNodes){ StringBuffer sb1 = new StringBuffer(); StringBuffer sb2 = new StringBuffer(); sb1.append(Thread.currentThread().getName()+"==加權輪詢--[當前權重]值的變化:"+printCurrentWeight(serverNodes)); // 有限權重總和可能發(fā)生變化 for(ServerNode serverNode : serverNodes){ totalWeight += serverNode.getEffectiveWeight(); } // 選出當前權重最大的節(jié)點 ServerNode tempNodeOfMaxWeight = serverNodes.get(0); for (ServerNode serverNode : serverNodes) { if (serverNode.isAvailable()) { serverNode.onInvokeSuccess();//提權 sb2.append(Thread.currentThread().getName()+"==[正常節(jié)點]:"+serverNode+"\n"); } else { serverNode.onInvokeFault();//降權 sb2.append(Thread.currentThread().getName()+"==[宕機節(jié)點]:"+serverNode+"\n"); } tempNodeOfMaxWeight = tempNodeOfMaxWeight.compareTo(serverNode) > 0 ? tempNodeOfMaxWeight : serverNode; } // 必須new個新的節(jié)點實例來保存信息,否則引用指向同一個堆實例,后面的set操作將會修改節(jié)點信息 nodeOfMaxWeight = new ServerNode(tempNodeOfMaxWeight.getIp(),tempNodeOfMaxWeight.getWeight(),tempNodeOfMaxWeight.isAvailable()); nodeOfMaxWeight.setEffectiveWeight(tempNodeOfMaxWeight.getEffectiveWeight()); nodeOfMaxWeight.setCurrentWeight(tempNodeOfMaxWeight.getCurrentWeight()); // 調整當前權重比:按權重(effectiveWeight)的比例進行調整,確保請求分發(fā)合理。 tempNodeOfMaxWeight.setCurrentWeight(tempNodeOfMaxWeight.getCurrentWeight() - totalWeight); sb1.append(" -> "+printCurrentWeight(serverNodes)); serverNodes.forEach(serverNode -> serverNode.setCurrentWeight(serverNode.getCurrentWeight()+serverNode.getEffectiveWeight())); sb1.append(" -> "+printCurrentWeight(serverNodes)); System.out.print(sb2); //所有節(jié)點的當前信息 System.out.println(sb1); //打印當前權重變化過程 } return nodeOfMaxWeight; } // 格式化打印信息 private String printCurrentWeight(List<ServerNode> serverNodes){ StringBuffer stringBuffer = new StringBuffer("["); serverNodes.forEach(node -> stringBuffer.append(node.getCurrentWeight()+",") ); return stringBuffer.substring(0, stringBuffer.length() - 1) + "]"; } // 并發(fā)測試:兩個線程循環(huán)獲取節(jié)點 public static void main(String[] args) throws InterruptedException { // 循環(huán)次數(shù) int loop = 18; new Thread(() -> { WeightedRoundRobinAvailable weightedRoundRobin1 = new WeightedRoundRobinAvailable(); for(int i=1;i<=loop;i++){ ServerNode serverNode = weightedRoundRobin1.selectNode(); System.out.println(Thread.currentThread().getName()+"==第"+i+"次輪詢選中[當前權重最大]的節(jié)點:" + serverNode + "\n"); } }).start(); // new Thread(() -> { WeightedRoundRobinAvailable weightedRoundRobin2 = new WeightedRoundRobinAvailable(); for(int i=1;i<=loop;i++){ ServerNode serverNode = weightedRoundRobin2.selectNode(); System.out.println(Thread.currentThread().getName()+"==第"+i+"次輪詢選中[當前權重最大]的節(jié)點:" + serverNode + "\n"); } }).start(); //main 線程睡了一下,再偷偷把 所有宕機 拉起來:模擬服務器恢復正常 Thread.sleep(5); for (ServerNode serverNode:serverNodes){ if(!serverNode.isAvailable()) serverNode.setIsAvailable(true); } } }
3.分析結果
執(zhí)行結果:將執(zhí)行結果的前中后四次抽出來分析
Thread-0==[正常節(jié)點]:{ip='192.168.1.101', weight=1, effectiveWeight=1, currentWeight=1, isAvailable=true}
Thread-0==[宕機節(jié)點]:{ip='192.168.1.102', weight=3, effectiveWeight=2, currentWeight=3, isAvailable=false}
Thread-0==[正常節(jié)點]:{ip='192.168.1.103', weight=2, effectiveWeight=2, currentWeight=2, isAvailable=true}
Thread-0==加權輪詢--[當前權重]值的變化:[1,3,2] -> [1,-3,2] -> [2,-1,4]
Thread-0==第1次輪詢選中[當前權重最大]的節(jié)點:{ip='192.168.1.102', weight=3, effectiveWeight=2, currentWeight=3, isAvailable=false}
……
Thread-1==[正常節(jié)點]:{ip='192.168.1.101', weight=1, effectiveWeight=1, currentWeight=6, isAvailable=true}
Thread-1==[宕機節(jié)點]:{ip='192.168.1.102', weight=3, effectiveWeight=-7, currentWeight=-21, isAvailable=false}
Thread-1==[正常節(jié)點]:{ip='192.168.1.103', weight=2, effectiveWeight=2, currentWeight=12, isAvailable=true}
Thread-1==加權輪詢--[當前權重]值的變化:[6,-21,12] -> [6,-21,15] -> [7,-28,17]
Thread-1==第5次輪詢選中[當前權重最大]的節(jié)點:{ip='192.168.1.103', weight=2, effectiveWeight=2, currentWeight=12, isAvailable=true}
……
Thread-0==[正常節(jié)點]:{ip='192.168.1.101', weight=1, effectiveWeight=1, currentWeight=13, isAvailable=true}
Thread-0==[正常節(jié)點]:{ip='192.168.1.102', weight=3, effectiveWeight=3, currentWeight=-19, isAvailable=true}
Thread-0==[正常節(jié)點]:{ip='192.168.1.103', weight=2, effectiveWeight=2, currentWeight=12, isAvailable=true}
Thread-0==加權輪詢--[當前權重]值的變化:[13,-19,12] -> [7,-19,12] -> [8,-16,14]
Thread-0==第15次輪詢選中[當前權重最大]的節(jié)點:{ip='192.168.1.101', weight=1, effectiveWeight=1, currentWeight=13, isAvailable=true}
……
Thread-1==[正常節(jié)點]:{ip='192.168.1.101', weight=1, effectiveWeight=1, currentWeight=2, isAvailable=true}
Thread-1==[正常節(jié)點]:{ip='192.168.1.102', weight=3, effectiveWeight=3, currentWeight=2, isAvailable=true}
Thread-1==[正常節(jié)點]:{ip='192.168.1.103', weight=2, effectiveWeight=2, currentWeight=2, isAvailable=true}
Thread-1==加權輪詢--[當前權重]值的變化:[2,2,2] -> [2,2,-4] -> [3,5,-2]
Thread-1==第18次輪詢選中[當前權重最大]的節(jié)點:{ip='192.168.1.103', weight=2, effectiveWeight=2, currentWeight=2, isAvailable=true}
分析
一開始權重最高的節(jié)點雖然是宕機了,但是還是會被選中并返回;
“有效權重總和” 和 “當前權重總和”都減少了1,因為設置輪詢到失敗節(jié)點,都會自減1;
到第5次輪詢時,當前權重已經(jīng)變成了[7,-28,17],可以看出宕機節(jié)點越往后當前權重越小,所以后面根本不會再選中宕機節(jié)點,雖然沒剔除故障節(jié)點,但卻起到不分配宕機節(jié)點;
到第15次輪詢時,有效權重已經(jīng)恢復起始值,當前權重變?yōu)閇8,-16,14],當前權重只能慢慢恢復,并不是節(jié)點一正常就立即恢復宕機過的節(jié)點,起到對故障節(jié)點的緩沖恢復(故障過的節(jié)點可能還存在問題);
最后1次輪詢時,因為沒有宕機節(jié)點,所以有效權重不變,當前權重已經(jīng)恢復[3,5,-2],如果再輪詢一次,那就會訪問到一開始故障的節(jié)點了。
4.結論
降權起到緩慢“剔除”宕機節(jié)點的效果;提權起到緩沖恢復宕機節(jié)點的效果。
對比上一篇文章可以看到:
當前權重(currentWeight):針對的是節(jié)點的選擇,受有效權重影響,起到緩慢“剔除”宕機節(jié)點和緩沖恢復宕機節(jié)點的效果,當前權重最高就會被選擇;
有效權重(effectiveWeight):針對的是權重的變化,也即是降權和提權,降權/提權只會直接操作有效權重;
權重(weight):針對的是存儲起始配置,限定有效權重的提權。
到此這篇關于Java實現(xiàn)平滑加權輪詢算法之降權和提權的文章就介紹到這了,更多相關Java平滑加權輪詢降權和提權內容請搜索腳本之家以前的文章或繼續(xù)瀏覽下面的相關文章希望大家以后多多支持腳本之家!
相關文章
spring框架下@value注解屬性static無法獲取值問題
這篇文章主要介紹了spring框架下@value注解屬性static無法獲取值問題,具有很好的參考價值,希望對大家有所幫助。如有錯誤或未考慮完全的地方,望不吝賜教2021-11-11詳解MyBatis-Puls中saveBatch批量添加慢的問題
本文主要介紹了詳解MyBatis-Puls中saveBatch批量添加慢的問題,文中通過示例代碼介紹的非常詳細,對大家的學習或者工作具有一定的參考學習價值,需要的朋友們下面隨著小編來一起學習學習吧2023-01-01springboot整合websocket實現(xiàn)群聊思路代碼詳解
通過springboot引入websocket,實現(xiàn)群聊,通過在線websocket測試進行展示。本文重點給大家介紹springboot整合websocket實現(xiàn)群聊功能,代碼超級簡單,感興趣的朋友跟隨小編一起學習吧2021-05-05java 文件大數(shù)據(jù)Excel下載實例代碼
這篇文章主要介紹了java 文件大數(shù)據(jù)Excel下載實例代碼的相關資料,需要的朋友可以參考下2017-04-04