一致性Hash算法
jopen
9年前
一致性Hash算法,比如负责均衡时候,连接到哪个节点,除了一般的随机算法,Round Robin算法外,还可以通过Hash映射到具体的ip上去
下面是golang版本的Hash一致性算法:
package main import ( "fmt" "hash/crc32" "sort" "strconv" "sync" ) const DEFAULT_REPLICAS = 160 type HashRing []uint32 func (c HashRing) Len() int { return len(c) } func (c HashRing) Less(i, j int) bool { return c[i] < c[j] } func (c HashRing) Swap(i, j int) { c[i], c[j] = c[j], c[i] } //Hash节点 ID IP Port HostName Weight type Node struct { Id int Ip string Port int HostName string Weight int } // NewNode(i, "172.18.1."+si, 8080, "host_"+si, 1) func NewNode(id int, ip string, port int, name string, weight int) *Node { return &Node{ Id: id, Ip: ip, Port: port, HostName: name, Weight: weight, } } //一致性 type Consistent struct { Nodes map[uint32]Node numReps int Resources map[int]bool ring HashRing //附带存储c.Nodes,方便排序之用 sync.RWMutex } func NewConsistent() *Consistent { return &Consistent{ Nodes: make(map[uint32]Node), numReps: DEFAULT_REPLICAS, Resources: make(map[int]bool), ring: HashRing{}, } } func (c *Consistent) Add(node *Node) bool { c.Lock() defer c.Unlock() if _, ok := c.Resources[node.Id]; ok { return false } count := c.numReps * node.Weight //权重高备份数据越多 for i := 0; i < count; i++ { str := c.joinStr(i, node) //多备份 c.Nodes[c.hashStr(str)] = *(node) } c.Resources[node.Id] = true c.sortHashRing() return true } func (c *Consistent) sortHashRing() { c.ring = HashRing{} for k := range c.Nodes { //这样遍历map只能返回key,uint32的hash key c.ring = append(c.ring, k) } sort.Sort(c.ring) } func (c *Consistent) joinStr(i int, node *Node) string { return node.Ip + "*" + strconv.Itoa(node.Weight) + "-" + strconv.Itoa(i) + "-" + strconv.Itoa(node.Id) } //最核心都是key映射成ID的算法 // 高运算性能,低碰撞率Hash算法-----MurMurHash算法 :https://github.com/spaolacci/murmur3 func (c *Consistent) hashStr(key string) uint32 { return crc32.ChecksumIEEE([]byte(key)) } func (c *Consistent) Get(key string) Node { c.RLock() defer c.RUnlock() hash := c.hashStr(key) i := c.search(hash) return c.Nodes[c.ring[i]] } func (c *Consistent) search(hash uint32) int { i := sort.Search(len(c.ring), func(i int) bool { return c.ring[i] >= hash }) if i < len(c.ring) { if i == len(c.ring)-1 { return 0 } else { return i } } else { return len(c.ring) - 1 } } func (c *Consistent) Remove(node *Node) { c.Lock() defer c.Unlock() if _, ok := c.Resources[node.Id]; !ok { return } delete(c.Resources, node.Id) //删除标志 count := c.numReps * node.Weight for i := 0; i < count; i++ { str := c.joinStr(i, node) delete(c.Nodes, c.hashStr(str)) //删除内容 } c.sortHashRing() } func main() { cHashRing := NewConsistent() for i := 0; i < 10; i++ { si := fmt.Sprintf("%d", i) cHashRing.Add(NewNode(i, "172.18.1."+si, 8080, "host_"+si, 1)) } ipMap := make(map[string]int, 0) for i := 0; i < 1000; i++ { si := fmt.Sprintf("key%d", i) k := cHashRing.Get(si) if _, ok := ipMap[k.Ip]; ok { ipMap[k.Ip] += 1 } else { ipMap[k.Ip] = 1 } } for k, v := range ipMap { fmt.Println("Node IP:", k, " count:", v) } }
总结:最核心就是MurMurHash算法,怎么把Hash的key映射成具体的ID
来自: http://studygolang.com/wr?u=http%3a%2f%2fblog.csdn.net%2fxcltapestry%2farticle%2fdetails%2f43898807