很好用的压测工具 - Apache Bench 工具
今年公司开发者大会是线上的形式，按照惯例，为了服务的保障，需要对整个系统的性能做一个评估，临时抱佛脚，比较常用的工具有 jmeter 和 Apache Bench，最终在两者之间选择了 Apache Bench(简称 ab)，也就针对 ab 工具做了一些总结。
ab is a tool for benchmarking your Apache Hypertext Transfer Protocol (HTTP) server. It is designed to give you an impression
of how your current Apache installation performs. This especially shows you how many requests per second your Apache installation
is capable of serving.
ApacheBench 是 Apache 服务器自带的一个 web 压力测试工具，简称 ab。ab 又是一个行工具，对发起负载的本机要求很低，根据 ab 可以创建很多的并发访问线程，模拟多个访问者同时对某一 URL 地址进行访问，因此可以用来测试目标服务器的负载压力。总的来说 ab 工具小巧简单，上手学习较快，可以提供需要的基本性能指标，但是没有图形化结果，不能监控。
jmeter 和 ab 的比较
1、jmeter 是一次完整的请求和返回， 而 AB 只是发出去请求，并不对返回做处理，只是请求发送成功或者失败。 所以从准确性来说，Jmeter 更准确，而 AB 速度更快，可以用最少的机器资源产生更多的访问请求；
2、Jmeter 本身支持断言、可变参数和 CSV 数据集的输入，能设定更加灵活多变的的测试场景，而 AB 则不支持（暂时没想到）；
3、Jmeter 可以提供更加详细的统计结果数据，比如接口错误信息、单线程的请求时间等，而 AB 则不支持；
4、Jmeter 不支持精确时间的压测，比如压测 10 分钟，但是 AB 支持；
5、Jmeter 支持分布式的压测集群，且支持函数，AB 不支持；
6、软件自身耗费资源：Jmeter 由于比较重，且统计了很多结果数据，比 AB 耗时耗费资源多，AB 属于超轻量级，在开发测试过程中十分适合做单接口压测。
因为本次只针对单个接口做测试，手上刚好有空闲的 linux 机器，综合考虑就选择了 AB，废话不多说，下面就进行 AB 的使用做一些讲解。
官网针对 ab 的使用做了很详细的介绍，我们可以去查看官网地址：
ab [ -A auth-username:password ] [ -b windowsize ] [ -B local-address ] [ -c concurrency ] [ -C cookie-name=value ] [ -d ]
[ -e csv-file ] [ -E client-certificate file ] [ -f protocol ] [ -g gnuplot-file ] [ -h ] [ -H custom-header ] [ -i ] [ -k ]
[ -l ] [ -m HTTP-method ] [ -n requests ] [ -p POST-file ] [ -P proxy-auth-username:password ] [ -q ] [ -r ] [ -s timeout ]
[ -S ] [ -t timelimit ] [ -T content-type ] [ -u PUT-file ] [ -v verbosity] [ -V ] [ -w ] [ -x <table>-attributes ]
[ -X proxy[:port] ] [ -y <tr>-attributes ] [ -z <td>-attributes ] [ -Z ciphersuite ] [http[s]://]hostname[:port]/path
Supply BASIC Authentication credentials to the server. The username and password are separated by a single : and sent on the wire base64 encoded. The string is sent regardless of whether the server needs it (i.e., has sent an 401 authentication needed).
Add a Cookie: line to the request. The argument is typically in the form of a name=value pair. This field is repeatable.
Write a Comma separated value (CSV) file which contains for each percentage (from 1% to 100%) the time (in milliseconds) it took to serve that percentage of the requests. This is usually more useful than the 'gnuplot' file; as the results are already 'binned'.
When connecting to an SSL website, use the provided client certificate in PEM format to authenticate with the server. The file is expected to contain the client certificate, followed by intermediate certificates, followed by the private key. Available in 2.4.36 and later.
Specify SSL/TLS protocol (SSL2, SSL3, TLS1, TLS1.1, TLS1.2, or ALL). TLS1.1 and TLS1.2 support available in 2.4.4 and later.
Write all measured values out as a 'gnuplot' or TSV (Tab separate values) file. This file can easily be imported into packages like Gnuplot, IDL, Mathematica, Igor or even Excel. The labels are on the first line of the file.
Append extra headers to the request. The argument is typically in the form of a valid header line, containing a colon-separated field-value pair (i.e., "Accept-Encoding: zip/zop;8bit").
Enable the HTTP KeepAlive feature, i.e., perform multiple requests within one HTTP session. Default is no KeepAlive.
Do not report errors if the length of the responses is not constant. This can be useful for dynamic pages. Available in 2.4.7 and later.
Number of requests to perform for the benchmarking session. The default is to just perform a single request which usually leads to non-representative benchmarking results.
Supply BASIC Authentication credentials to a proxy en-route. The username and password are separated by a single : and sent on the wire base64 encoded. The string is sent regardless of whether the proxy needs it (i.e., has sent an 407 proxy authentication needed).
When processing more than 150 requests, ab outputs a progress count on stderr every 10% or 100 requests or so. The -q flag will suppress these messages.
Maximum number of seconds to wait before the socket times out. Default is 30 seconds. Available in 2.4.4 and later.
Do not display the median and standard deviation values, nor display the warning/error messages when the average and median are more than one or two times the standard deviation apart. And default to the min/avg/max values. (legacy support).
Maximum number of seconds to spend for benchmarking. This implies a -n 50000 internally. Use this to benchmark the server within a fixed total amount of time. Per default there is no timelimit.
Content-type header to use for POST/PUT data, eg. application/x-www-form-urlencoded. Default is text/plain.
Set verbosity level - 4 and above prints information on headers, 3 and above prints response codes (404, 200, etc.), 2 and above prints warnings and info.
The value, if any, returned in the server HTTP header of the first successful response. This includes all characters in the header from beginning to the point a character with decimal value of 32 (most notably: a space or CR/LF) is detected.
The DNS or IP address given on the command line
The port to which ab is connecting. If no port is given on the command line, this will default to 80 for http and 443 for https.
The protocol parameters negotiated between the client and server. This will only be printed if SSL is used.
The request URI parsed from the command line string.
This is the size in bytes of the first successfully returned document. If the document length changes during testing, the response is considered an error.
The number of concurrent clients used during the test
Time taken for tests
This is the time taken from the moment the first socket connection is created to the moment the last response is received
The number of successful responses received
The number of requests that were considered a failure. If the number is greater than zero, another line will be printed showing the number of requests that failed due to connecting, reading, incorrect content length, or exceptions.
The number of errors that failed during write (broken pipe).
The number of responses that were not in the 200 series of response codes. If all responses were 200, this field is not printed.
The number of connections that resulted in Keep-Alive requests
Total body sent
If configured to send data as part of the test, this is the total number of bytes sent during the tests. This field is omitted if the test did not include a body to send.
The total number of bytes received from the server. This number is essentially the number of bytes sent over the wire.
The total number of document bytes received from the server. This number excludes bytes received in HTTP headers
Requests per second
This is the number of requests per second. This value is the result of dividing the number of requests by the total time taken
Time per request
The average time spent per request. The first value is calculated with the formula concurrency * timetaken * 1000 / done while the second value is calculated with the formula timetaken * 1000 / done
The rate of transfer as calculated by the formula totalread / 1024 / timetaken
从上面可以看到 ab 支持的参数很多，但其实我们一般来说只有-n（发送的请求总数）和 -c（请求的并发线程数）的参数是必要的。
从图上的列子我们可以看出，本次是启动了 200 个线程并发，总共跑了 2000 次（为了避免泄露信息安全，对接口做了隐私处理）。
Requests for second 为 722.77,即每秒钟处理 722.77 个请求；
Time per request 为 276.714,即请求的平均耗时为 276.714ms；
另外还可以 99 线，98 线，95 线等的耗时指标；
特别要注意，NON-2xx responses 为 2000，而请求总共 2000，则表示 2000 个请求返回的全部都是非 2000，需要考虑请求返回是否都是错误的；
版权声明: 本文为 InfoQ 作者【恒生LIGHT云社区】的原创文章。