你可能写了个假异步,并不能提高请求线程池的吞吐量
Posted 秋荷雨翔的博客
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了你可能写了个假异步,并不能提高请求线程池的吞吐量相关的知识,希望对你有一定的参考价值。
不知道用什么词形容,就叫它假异步吧。
写异步方法,async 和 await 要一路写到底,否则就是假异步,并不能提高请求线程池的吞吐量。
真正的异步,我的理解是这样的:比如调用一个查询接口,在当前线程,把SQL扔给数据库,当前线程释放,去干别的事情,数据库查询完了,通知我,我再在另一个线程里(也可能是刚才释放的那个线程,也可能不是)拿查询结果,返回给客户端,数据库查询比较耗时,数据库查询的时候,对线程是0占用。
HttpUtil.HttpGet方法:
/// <summary> /// HttpGet /// </summary> /// <param name="url">url路径名称</param> /// <param name="cookie">cookie</param> public static string HttpGet(string url, CookieContainer cookie = null, WebHeaderCollection headers = null) { try { // 设置参数 HttpWebRequest request = WebRequest.Create(url) as HttpWebRequest; request.CookieContainer = cookie; request.Method = "GET"; request.ContentType = "text/plain;charset=utf-8"; if (headers != null) { foreach (string key in headers.Keys) { request.Headers.Add(key, headers[key]); } } //发送请求并获取相应回应数据 HttpWebResponse response = request.GetResponse() as HttpWebResponse; //直到request.GetResponse()程序才开始向目标网页发送Post请求 Stream instream = response.GetResponseStream(); StreamReader sr = new StreamReader(instream, Encoding.UTF8); //返回结果网页(html)代码 string content = sr.ReadToEnd(); instream.Close(); return content; } catch (Exception ex) { LogUtil.Error(ex); return string.Empty; } }
HttpUtil.HttpGetAsync方法:
/// <summary> /// HttpGetAsync /// </summary> /// <param name="url">url路径名称</param> /// <param name="cookie">cookie</param> public static async Task<string> HttpGetAsync(string url, CookieContainer cookie = null, WebHeaderCollection headers = null) { try { // 设置参数 HttpWebRequest request = WebRequest.Create(url) as HttpWebRequest; request.CookieContainer = cookie; request.Method = "GET"; request.ContentType = "text/plain;charset=utf-8"; if (headers != null) { foreach (string key in headers.Keys) { request.Headers.Add(key, headers[key]); } } //发送请求并获取相应回应数据 HttpWebResponse response = await request.GetResponseAsync() as HttpWebResponse; //直到request.GetResponse()程序才开始向目标网页发送Post请求 Stream instream = response.GetResponseStream(); StreamReader sr = new StreamReader(instream, Encoding.UTF8); //返回结果网页(html)代码 string content = sr.ReadToEnd(); instream.Close(); return content; } catch (Exception ex) { LogUtil.Error(ex); return string.Empty; } }
测试代码:
/// <summary> /// 测试1 /// </summary> private async void button1_Click(object sender, EventArgs e) { //task是自己写的独立线程池,为了防止测试过程对CLR线程池和异步线程池造成干扰 _task.Run(() => { Thread.Sleep(200); int workerThreads1, completionPortThreads1, workerThreads2, completionPortThreads2; ThreadPool.GetMaxThreads(out workerThreads1, out completionPortThreads1); ThreadPool.GetAvailableThreads(out workerThreads2, out completionPortThreads2); Log("假异步 已使用辅助线程:" + (workerThreads1 - workerThreads2) + ", 已使用异步线程:" + (completionPortThreads1 - completionPortThreads2)); }); string str = await GetDataAsync(); } /// <summary> /// 测试2 /// </summary> private async void button2_Click(object sender, EventArgs e) { //task是自己写的独立线程池,为了防止测试过程对CLR线程池和异步线程池造成干扰 _task.Run(() => { Thread.Sleep(200); int workerThreads1, completionPortThreads1, workerThreads2, completionPortThreads2; ThreadPool.GetMaxThreads(out workerThreads1, out completionPortThreads1); ThreadPool.GetAvailableThreads(out workerThreads2, out completionPortThreads2); Log("真异步 已使用辅助线程:" + (workerThreads1 - workerThreads2) + ", 已使用异步线程:" + (completionPortThreads1 - completionPortThreads2)); }); string str = await GetDataAsync2(); } /// <summary> /// 假异步 /// </summary> private async Task<string> GetDataAsync() { return await Task.Run<string>(() => { //接口耗时大约1秒 return HttpUtil.HttpGet("http://localhost:8500/api/test/TestGet?val=1", null, null); }); } /// <summary> /// 真异步 /// </summary> /// <returns></returns> private async Task<string> GetDataAsync2() { //接口耗时大约1秒 return await HttpUtil.HttpGetAsync("http://localhost:8500/api/test/TestGet?val=1", null, null); }
测试截图:
我想知道 WebRequest 类的 GetResponseAsync 方法是怎么实现的,反正使用.NET提供的类,我是无法实现这样的方法。
这篇随笔如果有错误,请指正,我只是抛砖引玉,提出疑问或者说假设。
async await 要一路写到底啊,重构压力真大。
如果在异步方法中调用同步方法,又不想占用CLR线程池的线程,可以把耗时操作放到自己写的独立线程池中,例如:
/// <summary> /// 测试GET请求 /// </summary> /// <param name="val">测试参数</param> [HttpGet] [Route("TestGet")] [SwaggerResponse(HttpStatusCode.OK, "返回JSON", typeof(JsonListResult<TestGetResult>))] public async Task<HttpResponseMessage> TestGetAsync(string val) { var task = TaskHelper.RequestTask.Run(() => //使用自己写的独立线程池 { List<TestGetResult> list = new List<TestGetResult>(); Thread.Sleep(5000); //模拟耗时 for (int i = 1; i <= 10; i++) { TestGetResult item = new TestGetResult(); item.testValue1 = i.ToString(); item.testValue2 = i; item.testValue3 = "这是传入参数:" + val; list.Add(item); } var jsonResult = new JsonListResult<TestGetResult>(list, list.Count); return ApiHelper.ToJson(jsonResult); }); return await task; }
以上是关于你可能写了个假异步,并不能提高请求线程池的吞吐量的主要内容,如果未能解决你的问题,请参考以下文章