Using async-profiler to Optimize CPU Usage in a Dynamic QPS Test Case

I have long heard about the powerful capabilities of CPU flame graphs and various flame‑graph tools, and today I finally started trying a CPU flame‑graph generation tool.

Unfortunately, the flame‑graph plugin bundled with IntelliJ could not be used for various reasons, so I turned to the async‑profiler analysis tool as a replacement.

While testing random‑number performance, I used a dynamic QPS model case, learned how to use async‑profiler, and unexpectedly discovered an area for performance optimization that reduced CPU usage by 0.1%, marking my first result.

async-profiler

The installation and usage guide for this tool can be found online; I recommend checking the GitHub repository Wiki for details.

Case code

Below is the case code that uses a dynamic QPS model.

<code style="padding: 16px; color: #333; display: -webkit-box; font-family: Operator Mono, Consolas, Monaco, Menlo, monospace; font-size: 12px"><span style="line-height: 26px"><span style="font-weight: bold; line-height: 26px">class</span> <span style="color: #458; font-weight: bold; line-height: 26px">T</span> <span style="font-weight: bold; line-height: 26px">extends</span> <span style="color: #458; font-weight: bold; line-height: 26px">SourceCode</span> {</span><br/><br/>    <span style="font-weight: bold; line-height: 26px">static</span> <span style="font-weight: bold; line-height: 26px">void</span> main(String[] args) {<br/>        <span style="font-weight: bold; line-height: 26px">def</span> total = <span style="color: #008080; line-height: 26px">1000</span>_0000<br/>        <span style="font-weight: bold; line-height: 26px">def</span> index = <span style="font-weight: bold; line-height: 26px">new</span> AtomicInteger()<br/>        <span style="font-weight: bold; line-height: 26px">int</span> i = <span style="color: #008080; line-height: 26px">0</span><br/>        <span style="font-weight: bold; line-height: 26px">def</span> test = {<br/>            i++ % total<br/>            <span style="color: #998; font-style: italic; line-height: 26px">//            index.getAndIncrement() % total</span><br/>            getRandomInt(total)<br/>            sleep(<span style="color: #008080; line-height: 26px">0.01</span>)<br/>        }<br/>        <span style="font-weight: bold; line-height: 26px">new</span> FunQpsConcurrent(test, <span style="color: #d14; line-height: 26px">"测试随机性能"</span>).start()<br/>    }<br/>}<br/><br/></code>

The method that executes the task is

com.okcoin.hickwall.presses.funtester.frame.execute.FunQpsConcurrent#start

and its code is shown below:

<code style="padding: 16px; color: #333; display: -webkit-box; font-family: Operator Mono, Consolas, Monaco, Menlo, monospace; font-size: 12px">    <span style="line-height: 26px"><span style="font-weight: bold; line-height: 26px">void</span> <span style="color: #900; font-weight: bold; line-height: 26px">start</span><span style="line-height: 26px">()</span> </span>{<br/>        <span style="font-weight: bold; line-height: 26px">if</span> (executor == <span style="font-weight: bold; line-height: 26px">null</span>) executor = ThreadPoolUtil.createCachePool(Constant.THREADPOOL_MAX, <span style="color: #d14; line-height: 26px">"Q"</span>)<br/>        <span style="font-weight: bold; line-height: 26px">if</span> (Common.PERF_PLATFORM) controller = <span style="font-weight: bold; line-height: 26px">new</span> RedisController(<span style="font-weight: bold; line-height: 26px">this</span>)<br/>        <span style="font-weight: bold; line-height: 26px">if</span> (controller == <span style="font-weight: bold; line-height: 26px">null</span>) controller = <span style="font-weight: bold; line-height: 26px">new</span> FunTester();<br/>        <span style="font-weight: bold; line-height: 26px">new</span> Thread(controller, <span style="color: #d14; line-height: 26px">"receiver"</span>).start();<br/>        <span style="font-weight: bold; line-height: 26px">while</span> (key) {<br/>            ThreadPoolUtil.executeTask(executor, qps, produce, total, name)<br/>        }<br/>        stop()<br/>    }<br/><br/></code>

Optimization Process

The entire main thread spends most of its time in the while loop. I first generated a flame graph of the main thread, shown below:

From the pre‑optimization flame graph, the method

com.okcoin.hickwall.presses.funtester.frame.execute.ThreadPoolUtil#executeTask

consumes 0.53% CPU, while the getSecond method uses the most CPU because it creates a Calendar object. The relevant code is:

<code style="padding: 16px; color: #333; display: -webkit-box; font-family: Operator Mono, Consolas, Monaco, Menlo, monospace; font-size: 12px">        <span style="font-weight: bold; line-height: 26px">if</span> (Time.getSecond() % COUNT_INTERVAL == <span style="color: #008080; line-height: 26px">0</span>) {<br/>            <span style="font-weight: bold; line-height: 26px">int</span> real = total.sumThenReset() / COUNT_INTERVAL as <span style="font-weight: bold; line-height: 26px">int</span><br/>            def active = executor.getActiveCount()<br/>            def count = active == <span style="color: #008080; line-height: 26px">0</span> ? <span style="color: #008080; line-height: 26px">1</span> : active<br/>            log.info(<span style="color: #d14; line-height: 26px">"{} design QPS:{},actual QPS:{} active thread:{} per thread efficiency:{}"</span>, name, qps, real, active, real / count as <span style="font-weight: bold; line-height: 26px">int</span>)<br/>        }<br/></code>

The original intention was to output design QPS, actual QPS, and active thread count every few seconds. I suspected that using a timestamp check would be faster, so I replaced the code as follows:

<code style="padding: 16px; color: #333; display: -webkit-box; font-family: Operator Mono, Consolas, Monaco, Menlo, monospace; font-size: 12px">        <span style="font-weight: bold; line-height: 26px">if</span> (SourceCode.getMark() % COUNT_INTERVAL == <span style="color: #008080; line-height: 26px">0</span>) {<br/>            <span style="font-weight: bold; line-height: 26px">int</span> real = total.sumThenReset() / COUNT_INTERVAL as <span style="font-weight: bold; line-height: 26px">int</span><br/>            def active = executor.getActiveCount()<br/>            def count = active == <span style="color: #008080; line-height: 26px">0</span> ? <span style="color: #008080; line-height: 26px">1</span> : active<br/>            log.info(<span style="color: #d14; line-height: 26px">"{} design QPS:{},actual QPS:{} active thread:{} per thread efficiency:{}"</span>, name, qps, real, active, real / count as <span style="font-weight: bold; line-height: 26px">int</span>)<br/>        }<br/></code>

After rebuilding and running the test, I captured another flame graph for the main thread, shown below:

Post‑optimization, the CPU usage of

com.okcoin.hickwall.presses.funtester.frame.execute.ThreadPoolUtil#executeTask

dropped to 0.29%, while the

com.okcoin.hickwall.presses.funtester.frame.execute.FunQpsConcurrent#start

method now consumes 0.44% CPU, a reduction of 0.09% compared with the original 0.53%.

Rounded, the overall improvement is about 0.1%, which I consider a successful optimization. I also noticed that most of the remaining CPU time is spent in the sleep method, suggesting that the earlier conclusions about random‑number performance may need revisiting.

FunTester原创专题推荐~ 接口功能测试专题 性能测试专题 Groovy专题 Java、Groovy、Go、Python 单测&白盒 FunTester社群风采 测试理论鸡汤 FunTester视频专题 案例分享：方案、BUG、爬虫 UI自动化专题 测试工具专题 -- By FunTester