JMeter 4.0：响应时间统计插件推荐及性能指标问询

Hey folks! Let's dive into your JMeter 4.0 questions—first, I'll share the best plugins and reports to track requests with response times under thresholds like 200ms or 500ms, then break down your performance questions.

Recommended Plugins/Reports for Counting Requests Below Specific Response Times #

These tools will make it easy to track how many of your requests hit your response time targets:

• Filter Results Tool (from JMeter Plugins Standard Set)
  This is my go-to for filtering results by response time. You can use it in two ways:

  • GUI Mode: Install via the JMeter Plugins Manager, then head to Tools > Filter Results. Set the condition -RT <=200 (for 200ms) and point it at your test's .jtl result file. It'll generate a filtered file with only the requests that meet the threshold—just check the total count in the output.
  • Command Line: Run a command like this to automate filtering:

    java -jar JMeterPlugins-Standard.jar --tool FilterResults --input-file result.jtl --output-file filtered-200ms.jtl --include-label-regex ".*" --response-time-over 0 --response-time-under 200
    

    The filtered file's request count is your number of fast requests.

• Response Time Distribution Plugin
  Also part of the Standard Plugins Set, this listener breaks down response times into configurable intervals (you can set 0-200ms, 200-500ms, etc.). It shows both the absolute count and percentage of requests in each bucket—perfect for getting your "under 200ms" stats at a glance during or after a test.

• Enhanced HTML Report (Built-in + Plugins)
  JMeter 4.0's built-in HTML report generator can be boosted with plugins to add threshold stats. Run your test with jmeter -n -t testplan.jmx -l result.jtl -e -o report to generate the report, then customize it to include a section for requests under specific response times. Alternatively, use the enhanced Summary Report listener to calculate percentages manually by dividing fast request counts by total requests.

Answers to Your Performance Questions #

1. What QPS will keep 90% of response times under 200ms? #

There’s no universal answer here—it all depends on your system’s capacity (CPU, memory, database speed, network latency, etc.). Here’s how to find your specific number:

• Start with a low QPS (e.g., 10 requests per second) using a Thread Group or Stepping Thread Group (plugin) to control concurrency.
• Run the test long enough for the system to stabilize (5-10 minutes minimum) and monitor the 90th percentile (p90) response time with the Percentile Graph listener.
• If p90 stays below 200ms, increment the QPS (try 20, then 30, etc.) and repeat.
• Stop when the p90 just exceeds 200ms—the highest QPS that kept p90 under 200ms is your target.
• Pro tip: Monitor system resources (CPU, memory, disk I/O) during testing to spot bottlenecks that might spike response times.

2. How to get the count and percentage of requests with response time under 200ms? #

You’ve got a few simple options:

• Response Time Distribution: This listener lets you define a 0-200ms interval. It’ll show you the exact count of requests in that bucket and its percentage of total requests—no manual math needed.
• Filter Results Tool: After generating the filtered .jtl file (only requests ≤200ms), open it in JMeter’s Summary Report to get the count. Divide that by the total request count from your original test to get the percentage.
• Script the .jtl File: For automation, use a quick Python or bash script to parse the .jtl CSV file. Count rows where the elapsed column (response time) is ≤200, then divide by the total row count minus 1 (to exclude the header) for the percentage.

内容的提问来源于stack exchange，提问作者Dariusz