I have a web service that receives data in JSON format, processes the data, and then returns the result to the requester.
I want to measure the request, response, and total time using cURL.
My example request looks like:
curl -X POST -d @file server:portand I currently measure this using the time command in Linux:
time curl -X POST -d @file server:portThe time command only measures total time, though - which isn't quite what I am looking for.
Is there any way to measure request and response times using cURL?
15 Answers
From this brilliant blog post...
cURL supports formatted output for the details of the request (see the cURL manpage for details, under -w, –write-out <format>). For our purposes we’ll focus just on the timing details that are provided. Times below are in seconds.
Create a new file,
curl-format.txt, and paste in:time_namelookup: %{time_namelookup}s\n time_connect: %{time_connect}s\n time_appconnect: %{time_appconnect}s\n time_pretransfer: %{time_pretransfer}s\n time_redirect: %{time_redirect}s\n time_starttransfer: %{time_starttransfer}s\n ----------\n time_total: %{time_total}s\nMake a request:
curl -w "@curl-format.txt" -o /dev/null -s ""Or on Windows, it's...
curl -w "@curl-format.txt" -o NUL -s ""
What this does:
-w "@curl-format.txt" tells cURL to use our format file-o /dev/null redirects the output of the request to /dev/null-stells cURL not to show a progress meter"" is
the URL we are requesting. Use quotes particularly if your URL has "&" query string parameters
And here is what you get back:
time_namelookup: 0.001s time_connect: 0.037s time_appconnect: 0.000s time_pretransfer: 0.037s time_redirect: 0.000s
time_starttransfer: 0.092s ---------- time_total: 0.164sI have not yet seen an option to output the results in microseconds, but if you're aware of one, post in the comments below.
Make a Linux/Mac shortcut (alias)
alias curltime="curl -w \"@$HOME/.curl-format.txt\" -o /dev/null -s "Then you can simply call...
curltime wordpress.orgThanks to commenter Pete Doyle!
Make a Linux/Mac stand-alone script
This script does not require a separate .txt file to contain the formatting.
Create a new file, curltime, somewhere in your executable path, and paste in:
#!/bin/bash
curl -w @- -o /dev/null -s "$@" <<'EOF' time_namelookup: %{time_namelookup}\n time_connect: %{time_connect}\n time_appconnect: %{time_appconnect}\n time_pretransfer: %{time_pretransfer}\n time_redirect: %{time_redirect}\n time_starttransfer: %{time_starttransfer}\n ----------\n time_total: %{time_total}\n
EOFThen call it the same way as the alias:
curltime wordpress.orgMake a Windows shortcut (aka BAT file)
Create a new text file called curltime.bat in the same folder as curl.exe and curl-format.txt, and paste in the following line:
curl -w "@%~dp0curl-format.txt" -o NUL -s %*Then from the command line you can simply call:
curltime wordpress.org(Make sure the folder is listed in your Windows PATH variable to be able to use the command from any folder.)
Here is the answer:
curl -X POST -d @file server:port -w %{time_connect}:%{time_starttransfer}:%{time_total}All of the variables used with -w can be found in man curl.
Option 1: to measure total time:
curl -o /dev/null -s -w 'Total: %{time_total}s\n' Sample output:
Total: 0.441094sOption 2: to get time to establish connection, time to first byte (TTFB) and total time:
curl -o /dev/null -s -w 'Establish Connection: %{time_connect}s\nTTFB: %{time_starttransfer}s\nTotal: %{time_total}s\n' Sample output:
Establish Connection: 0.020033s
TTFB: 0.417907s
Total: 0.435486sRef: Get response time with curl
0The following is inspired by Simon's answer. It's self-contained (doesn't require a separate format file), which makes it great for inclusion into .bashrc.
curl_time() { curl -so /dev/null -w "\ namelookup: %{time_namelookup}s\n\ connect: %{time_connect}s\n\ appconnect: %{time_appconnect}s\n\ pretransfer: %{time_pretransfer}s\n\ redirect: %{time_redirect}s\n\
starttransfer: %{time_starttransfer}s\n\
-------------------------\n\ total: %{time_total}s\n" "$@"
}Futhermore, it should work with all arguments that curl normally takes, since the "$@" just passes them through. For example, you can do:
curl_time -X POST -H "Content-Type: application/json" -d '{"key": "val"}' Output:
namelookup: 0,125000s connect: 0,250000s appconnect: 0,609000s pretransfer: 0,609000s redirect: 0,000000s
starttransfer: 0,719000s
------------------------- total: 0,719000sA shortcut you can add to your .bashrc etc, based on other answers here:
function perf { curl -o /dev/null -s -w "%{time_connect} + %{time_starttransfer} = %{time_total}\n" "$1"
}Usage:
> perf stackoverflow.com
0.521 + 0.686 = 1.290If you want to analyze or summarize the latency you can try apache bench:
ab -n [number of samples] [url]For example:
ab -n 100 It will show:
This is ApacheBench, Version 2.3 <$Revision: 1757674 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd,
Licensed to The Apache Software Foundation,
Benchmarking (be patient).....done
Server Software: gws
Server Hostname:
Server Port: 80
Document Path: /
Document Length: 12419 bytes
Concurrency Level: 1
Time taken for tests: 10.700 seconds
Complete requests: 100
Failed requests: 97 (Connect: 0, Receive: 0, Length: 97, Exceptions: 0)
Total transferred: 1331107 bytes
HTML transferred: 1268293 bytes
Requests per second: 9.35 [#/sec] (mean)
Time per request: 107.004 [ms] (mean)
Time per request: 107.004 [ms] (mean, across all concurrent requests)
Transfer rate: 121.48 [Kbytes/sec] received
Connection Times (ms) min mean[+/-sd] median max
Connect: 20 22 0.8 22 26
Processing: 59 85 108.7 68 911
Waiting: 59 85 108.7 67 910
Total: 80 107 108.8 90 932
Percentage of the requests served within a certain time (ms) 50% 90 66% 91 75% 93 80% 95 90% 105 95% 111 98% 773 99% 932 100% 932 (longest request)Another way is configuring ~/.curlrc like this
-w "\n\n==== cURL measurements stats ====\ntotal: %{time_total} seconds \nsize: %{size_download} bytes \ndnslookup: %{time_namelookup} seconds \nconnect: %{time_connect} seconds \nappconnect: %{time_appconnect} seconds \nredirect: %{time_redirect} seconds \npretransfer: %{time_pretransfer} seconds \nstarttransfer: %{time_starttransfer} seconds \ndownloadspeed: %{speed_download} byte/sec \nuploadspeed: %{speed_upload} byte/sec \n\n"So the output of curl is
❯❯ curl -I
HTTP/2 301
location:
content-type: text/html; charset=UTF-8
date: Mon, 04 Mar 2019 08:02:43 GMT
expires: Wed, 03 Apr 2019 08:02:43 GMT
cache-control: public, max-age=2592000
server: gws
content-length: 220
x-xss-protection: 1; mode=block
x-frame-options: SAMEORIGIN
alt-svc: quic=":443"; ma=2592000; v="44,43,39"
==== cURL measurements stats ====
total: 0.211117 seconds
size: 0 bytes
dnslookup: 0.067179 seconds
connect: 0.098817 seconds
appconnect: 0.176232 seconds
redirect: 0.000000 seconds
pretransfer: 0.176438 seconds
starttransfer: 0.209634 seconds
downloadspeed: 0.000 byte/sec
uploadspeed: 0.000 byte/secYou can use curl -v --trace-time to add timestamps to the trace/verbose output. This must be done in verbose mode or trace mode to do anything.
Another option that is perhaps the simplest one in terms of the command line is adding the built-in --trace-time option:
curl -X POST -d @file server:port --trace-timeEven though it technically does not output the timings of the various steps as requested by the OP, it does display the timestamps for all steps of the request as shown below. Using this, you can (fairly easily) calculate how long each step has taken.
$ curl --trace-time -v -o /dev/null
13:29:11.148734 * Rebuilt URL to: % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 013:29:11.149958 * Trying 172.217.20.36...
13:29:11.149993 * TCP_NODELAY set
13:29:11.163177 * Connected to (172.217.20.36) port 443 (#0)
13:29:11.164768 * ALPN, offering h2
13:29:11.164804 * ALPN, offering http/1.1
13:29:11.164833 * successfully set certificate verify locations:
13:29:11.164863 * CAfile: none CApath: /etc/ssl/certs
13:29:11.165046 } [5 bytes data]
13:29:11.165099 * (304) (OUT), TLS handshake, Client hello (1):
13:29:11.165128 } [512 bytes data]
13:29:11.189518 * (304) (IN), TLS handshake, Server hello (2):
13:29:11.189537 { [100 bytes data]
13:29:11.189628 * TLSv1.2 (IN), TLS handshake, Certificate (11):
13:29:11.189658 { [2104 bytes data]
13:29:11.190243 * TLSv1.2 (IN), TLS handshake, Server key exchange (12):
13:29:11.190277 { [115 bytes data]
13:29:11.190507 * TLSv1.2 (IN), TLS handshake, Server finished (14):
13:29:11.190539 { [4 bytes data]
13:29:11.190770 * TLSv1.2 (OUT), TLS handshake, Client key exchange (16):
13:29:11.190797 } [37 bytes data]
13:29:11.190890 * TLSv1.2 (OUT), TLS change cipher, Client hello (1):
13:29:11.190915 } [1 bytes data]
13:29:11.191023 * TLSv1.2 (OUT), TLS handshake, Finished (20):
13:29:11.191053 } [16 bytes data]
13:29:11.204324 * TLSv1.2 (IN), TLS handshake, Finished (20):
13:29:11.204358 { [16 bytes data]
13:29:11.204417 * SSL connection using TLSv1.2 / ECDHE-ECDSA-CHACHA20-POLY1305
13:29:11.204451 * ALPN, server accepted to use h2
13:29:11.204483 * Server certificate:
13:29:11.204520 * subject: C=US; ST=California; L=Mountain View; O=Google LLC; CN=
13:29:11.204555 * start date: Oct 2 07:29:00 2018 GMT
13:29:11.204585 * expire date: Dec 25 07:29:00 2018 GMT
13:29:11.204623 * subjectAltName: host "" matched cert's ""
13:29:11.204663 * issuer: C=US; O=Google Trust Services; CN=Google Internet Authority G3
13:29:11.204701 * SSL certificate verify ok.
13:29:11.204754 * Using HTTP2, server supports multi-use
13:29:11.204795 * Connection state changed (HTTP/2 confirmed)
13:29:11.204840 * Copying HTTP/2 data in stream buffer to connection buffer after upgrade: len=0
13:29:11.204881 } [5 bytes data]
13:29:11.204983 * Using Stream ID: 1 (easy handle 0x55846ef24520)
13:29:11.205034 } [5 bytes data]
13:29:11.205104 > GET / HTTP/2
13:29:11.205104 > Host:
13:29:11.205104 > User-Agent: curl/7.61.0
13:29:11.205104 > Accept: */*
13:29:11.205104 >
13:29:11.218116 { [5 bytes data]
13:29:11.218173 * Connection state changed (MAX_CONCURRENT_STREAMS == 100)!
13:29:11.218211 } [5 bytes data]
13:29:11.251936 < HTTP/2 200
13:29:11.251962 < date: Fri, 19 Oct 2018 10:29:11 GMT
13:29:11.251998 < expires: -1
13:29:11.252046 < cache-control: private, max-age=0
13:29:11.252085 < content-type: text/html; charset=ISO-8859-1
13:29:11.252119 < p3p: CP="This is not a P3P policy! See for more info."
13:29:11.252160 < server: gws
13:29:11.252198 < x-xss-protection: 1; mode=block
13:29:11.252228 < x-frame-options: SAMEORIGIN
13:29:11.252262 < set-cookie: 1P_JAR=2018-10-19-10; expires=Sun, 18-Nov-2018 10:29:11 GMT; path=/; domain=.google.com
13:29:11.252297 < set-cookie: NID=141=pzXxp1jrJmLwFVl9bLMPFdGCtG8ySQKxB2rlDWgerrKJeXxfdmB1HhJ1UXzX-OaFQcnR1A9LKYxi__PWMigjMBQHmI3xkU53LI_TsYRbkMNJNdxs-caQQ7fEcDGE694S; expires=Sat, 20-Apr-2019 10:29:11 GMT; path=/; domain=.google.com; HttpOnly
13:29:11.252336 < alt-svc: quic=":443"; ma=2592000; v="44,43,39,35"
13:29:11.252368 < accept-ranges: none
13:29:11.252408 < vary: Accept-Encoding
13:29:11.252438 <
13:29:11.252473 { [5 bytes data]
100 12215 0 12215 0 0 112k 0 --:--:-- --:--:-- --:--:-- 112k
13:29:11.255674 * Connection #0 to host left intactHey is better than Apache Bench, has fewer issues with SSL
./hey -more
Summary: Total: 3.0960 secs Slowest: 1.6052 secs Fastest: 0.4063 secs Average: 0.6773 secs Requests/sec: 64.5992
Response time histogram: 0.406 [1] | 0.526 [142] |∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎ 0.646 [1] | 0.766 [6] |∎∎ 0.886 [0] | 1.006 [0] | 1.126 [0] | 1.246 [12] |∎∎∎ 1.365 [32] |∎∎∎∎∎∎∎∎∎ 1.485 [5] |∎ 1.605 [1] |
Latency distribution: 10% in 0.4265 secs 25% in 0.4505 secs 50% in 0.4838 secs 75% in 1.2181 secs 90% in 1.2869 secs 95% in 1.3384 secs 99% in 1.4085 secs
Details (average, fastest, slowest): DNS+dialup: 0.1150 secs, 0.0000 secs, 0.4849 secs DNS-lookup: 0.0032 secs, 0.0000 secs, 0.0319 secs req write: 0.0001 secs, 0.0000 secs, 0.0007 secs resp wait: 0.2068 secs, 0.1690 secs, 0.4906 secs resp read: 0.0117 secs, 0.0011 secs, 0.2375 secs
Status code distribution: [200] 200 responsesReferences
0I made a friendly formatter for sniffing curl requests to help with debugging ( see comments for usage ). It contains's every known output parameter you can write out in an easy to read format.
here is the string you can use with -w, contains all options that curl -w supports.
{"contentType":"%{content_type}","filenameEffective":"%{filename_effective}","ftpEntryPath":"%{ftp_entry_path}","httpCode":"%{http_code}","httpConnect":"%{http_connect}","httpVersion":"%{http_version}","localIp":"%{local_ip}","localPort":"%{local_port}","numConnects":"%{num_connects}","numRedirects":"%{num_redirects}","proxySslVerifyResult":"%{proxy_ssl_verify_result}","redirectUrl":"%{redirect_url}","remoteIp":"%{remote_ip}","remotePort":"%{remote_port}","scheme":"%{scheme}","size":{"download":"%{size_download}","header":"%{size_header}","request":"%{size_request}","upload":"%{size_upload}"},"speed":{"download":"%{speed_download}","upload":"%{speed_upload}"},"sslVerifyResult":"%{ssl_verify_result}","time":{"appconnect":"%{time_appconnect}","connect":"%{time_connect}","namelookup":"%{time_namelookup}","pretransfer":"%{time_pretransfer}","redirect":"%{time_redirect}","starttransfer":"%{time_starttransfer}","total":"%{time_total}"},"urlEffective":"%{url_effective}"}outputs JSON.
1Here's a Bash one-liner to hit the same server repeatedly:
for i in {1..1000}; do curl -s -o /dev/null -w "%{time_total}\n" doneThis is a modified version of Simons answer which makes the multi-lined output a single line. It also introduces the current timestamp so it's easier to follow each line of output.
Sample format fle$ cat time-format.txt
time_namelookup:%{time_namelookup} time_connect:%{time_connect} time_appconnect:%{time_appconnect} time_pretransfer:%{time_pretransfer} time_redirect:%{time_redirect} time_starttransfer:%{time_starttransfer} time_total:%{time_total}\n$ while [ 1 ];do echo -n "$(date) - " ; curl -w @time-format.txt -o /dev/null -s sleep 1; done | grep -v time_total:0Mon Dec 16 17:51:47 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.172 time_pretransfer:0.172 time_redirect:0.000 time_starttransfer:1.666 time_total:1.666
Mon Dec 16 17:51:50 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.175 time_pretransfer:0.175 time_redirect:0.000 time_starttransfer:3.794 time_total:3.795
Mon Dec 16 17:51:55 UTC 2019 - time_namelookup:0.004 time_connect:0.017 time_appconnect:0.175 time_pretransfer:0.175 time_redirect:0.000 time_starttransfer:1.971 time_total:1.971
Mon Dec 16 17:51:58 UTC 2019 - time_namelookup:0.004 time_connect:0.014 time_appconnect:0.173 time_pretransfer:0.173 time_redirect:0.000 time_starttransfer:1.161 time_total:1.161
Mon Dec 16 17:52:00 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.166 time_pretransfer:0.167 time_redirect:0.000 time_starttransfer:1.434 time_total:1.434
Mon Dec 16 17:52:02 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.177 time_pretransfer:0.177 time_redirect:0.000 time_starttransfer:5.119 time_total:5.119
Mon Dec 16 17:52:08 UTC 2019 - time_namelookup:0.004 time_connect:0.014 time_appconnect:0.172 time_pretransfer:0.172 time_redirect:0.000 time_starttransfer:30.185 time_total:30.185
Mon Dec 16 17:52:39 UTC 2019 - time_namelookup:0.004 time_connect:0.014 time_appconnect:0.164 time_pretransfer:0.164 time_redirect:0.000 time_starttransfer:30.175 time_total:30.176
Mon Dec 16 17:54:28 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:3.191 time_pretransfer:3.191 time_redirect:0.000 time_starttransfer:3.212 time_total:3.212
Mon Dec 16 17:56:08 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:1.184 time_pretransfer:1.184 time_redirect:0.000 time_starttransfer:1.215 time_total:1.215
Mon Dec 16 18:00:24 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.181 time_pretransfer:0.181 time_redirect:0.000 time_starttransfer:1.267 time_total:1.267I used the above to catch slow responses on the above endpoint.
Generic solution to measuring command execution time
Linux / Mac OS / Unix
On Linux you can measure the time taken for a command to execute with the time command
time curl Output:
ect...
</body>
</html>
real 0m0.565s
user 0m0.011s
sys 0m0.024sHere the overall time / response time is the real time.
This solution is generic and should work with any Unix command
time sleep 5Output:
real 0m5.001s
user 0m0.001s
sys 0m0.000sWindows Powershell
You can use Measure-Command
