1、线程池网络服务
:针对多线程网络服务模式的一些不足之处而提出的改进模式。
池是一个很重要的概念,其基本理念是:先创建一批资源,当有用户到来时,直接分配以创建好的资源,它的主要目的是减少系统在频繁创建资源时的开销。
实现原理:主服务线程创建既定数量的服务线程,当有客户端到来时,则从线程池中找出空闲的服务线程,为其服务,服务完毕后,线程不进行释放,重新放回线程池;若当前线程池已满,则将当前的客户端加入等待队列。
模型如下:
2、代码实现
(1)、utili.h
#include<unistd.h> #include<stdio.h> #include<string.h> #include<stdlib.h> #include<sys/socket.h> #include<netinet/in.h> #include<arpa/inet.h> #include<pthread.h> #define SERVER_PORT 8090 #define SERVER_IP "127.0.0.1" #define LISTEN_QUEUE 5 #define BUFFER_SIZE 255 #define CMD_SIZE 20 #define THREAD_POOL_NUM 5 typedef enum{ADD,SUB,MUL,DIV,MOD, QUIT}OPER_TYPE; typedef enum{IDEL, BUSY}THREAD_TAG; typedef struct OperStruct{ int op1; int op2; OPER_TYPE oper; }OperStruct;
(2)、ser.c
#include"../utili.h" typedef struct PoolStruct{ int sockConn; THREAD_TAG flag; }PoolStruct; typedef PoolStruct threadpool[THREAD_POOL_NUM]; threadpool pool; pthread_t tid[THREAD_POOL_NUM]; void* Thread_Handler(void *arg); void* Thread_Handler(void *arg){ int index = *(int *)arg; printf("[%d] thread start up.\n", index); OperStruct op; int result; while(1){ if(pool[index].flag == BUSY){ printf("[%d] thread start wroking.\n", index); int res = recv(pool[index].sockConn, &op, sizeof(op), 0); if(res == -1){ printf("recv data fail.\n"); continue; } if(op.oper == ADD){ result = op.op1 + op.op2; }else if(op.oper == SUB){ result = op.op1 - op.op2; }else if(op.oper == MUL){ result = op.op1 * op.op2; }else if(op.oper == DIV){ result = op.op1 / op.op2; }else if(op.oper == QUIT){ break; } res = send(pool[index].sockConn, &result, sizeof(result), 0); if(res == -1){ printf("send data fail.\n"); continue; } }else{ printf("[%d] thread sleep.\n",index); sleep(1); } } close(pool[index].sockConn); pthread_exit(0); } int main(void){ int sockSer = socket(AF_INET, SOCK_STREAM, 0); if(sockSer == -1){ perror("socket"); return -1; } struct sockaddr_in addrSer, addrCli; addrSer.sin_family = AF_INET; addrSer.sin_port = htons(SERVER_PORT); addrSer.sin_addr.s_addr = inet_addr(SERVER_IP); socklen_t len = sizeof(struct sockaddr); int res = bind(sockSer, (struct sockaddr*)&addrSer, len); if(res == -1){ perror("bind"); close(sockSer); return -1; } listen(sockSer, LISTEN_QUEUE); int i; for(i=0; i<THREAD_POOL_NUM; ++i){ pthread_create(&tid[i], NULL, Thread_Handler, &i); sleep(1); } for(i=0; i<THREAD_POOL_NUM; ++i){ pool[i].sockConn = 0; pool[i].flag = IDEL; } int sockConn; while(1){ printf("Server Wait Client Connect.......\n"); sockConn = accept(sockSer, (struct sockaddr*)&addrCli, &len); if(sockConn == -1){ printf("Server Accept Client Connect Fail.\n"); continue; }else{ printf("Server Accept Client Connect Success.\n"); printf("Client IP:>%s\n", inet_ntoa(addrCli.sin_addr)); printf("Client Port:>%d\n",ntohs(addrCli.sin_port)); } for(i=0; i<THREAD_POOL_NUM; ++i){ if(pool[i].flag == IDEL){ pool[i].flag = BUSY; pool[i].sockConn = sockConn; break; } } } close(sockSer); return 0; }
(3)、cli.c
#include"utili.h" void InputData(OperStruct *pt); void InputData(OperStruct *pt){ printf("please input op1 and op2 : "); scanf("%d %d", &(pt->op1), &(pt->op2)); } //Cli int main(void){ int sockCli = socket(AF_INET, SOCK_STREAM, 0); if(sockCli == -1){ perror("socket"); return -1; } struct sockaddr_in addrSer; addrSer.sin_family = AF_INET; addrSer.sin_port = htons(SERVER_PORT); addrSer.sin_addr.s_addr = inet_addr(SERVER_IP); socklen_t len = sizeof(struct sockaddr); int res = connect(sockCli, (struct sockaddr*)&addrSer, len); if(res == -1){ perror("connect"); close(sockCli); return -1; }else{ printf("Client Connect Server Success.\n"); } char cmd[2]; OperStruct op; int result; while(1){ printf("Please input operator : "); scanf("%s",cmd); if(strcmp(cmd, "+") == 0){ op.oper = ADD; InputData(&op); }else if(strcmp(cmd,"-") == 0){ op.oper = SUB; InputData(&op); }else if(strcmp(cmd,"*") == 0){ op.oper = MUL; InputData(&op); }else if(strcmp(cmd,"/") == 0){ op.oper = DIV; InputData(&op); }else if(strcmp(cmd, "quit") == 0){ op.oper = QUIT; }else{ printf("Cmd invalid.\n"); } res = send(sockCli, &op, sizeof(op), 0); if(res == -1){ printf("send data fail.\n"); continue; } if(op.oper == QUIT) break; res = recv(sockCli, &result, sizeof(result), 0); if(res == -1){ printf("recv data fail.\n"); continue; } printf("result = %d\n", result); } close(sockCli); return 0; }
运行结果
服务器端
客户端2
3、分析总结
(1)、其优点:性能高效
(2)、可能存在的问题:新用户如果在等待队列里耗时过长,会影响用户体验,针对此问题,改进方案如下:
a、动态创建新的服务线程,服务结束后,该线程加入线程池,这种改进的好处是,用户体验得到提升,潜在问题是,在长时间,大规模的并发用户状态下,线程会产生很多,最终会因为资源消耗过多,系统退出。
b、增加一个线程资源回收机制,当线程池的规模达到一定程度或满足某种既定规则时,会主动杀死一些线程,以达到系统稳定和用户体验之间折中。
模型分析
当有客户端来,有2种做法,i>、创建线程为其服务;ii>、加入等待队列;这2种都不太合适,采用折中法,有一个上限值,即就是规定一个创建线程的最大数,当来一个用户,还没达到线程最大数时,为其创建线程,若达到了,则加入等待队列;
对线程资源的回收:i>、立马回收,ii>、暂时不回收;当空闲的线程数达到某一下限值时,此时再将线程回收;