http-ping is a free software distributed under the Apache License 2.0.
This piece of software is similar to the usual ping networking utility but instead of working on top of ICMP, it works on top of HTTP/S.
Http-Ping is a small, free, easy-to-use command-line utility that probes a given URL and displays relevant statistics. It is similar to the popular ping utility, but works over HTTP/S instead of ICMP, and with a URL instead of a computer name/IP address. http-ping supports IPv6 addresses.
This software is written in Go, and should then benefit from the wide list of targets provided by Go.
This software has been reported to work well on:
- Linux: amd64, 386, arm64, arm
- FreeBSD: amd64, 386
- Windows: amd64, 386, arm64
- MacOS: amd64 (Intel Macs), arm64 (Apple Silicon)
Simply type http-ping -h to get the list of available commands
$ http-ping -h
An utility that evaluates the latency and throuput of HTTP/S requests
Usage:
http-ping [flags] target-URL
Flags:
-a, --audible-bell audible ; include a bell (ASCII 0x07) character in the outhroughput when any successful answer is received
--auth-password string authentication password
--auth-username string authentication username
--conn-target string force connection to be done with a specific IP:port (i.e. 127.0.0.1:8080)
--cookie string add one or more cookies, in the form name=value
-c, --count int define the number of request to be sent (default unlimited)
--detect-versions detect HTTP protocol versions available on target
--disable-compression the client will not request the remote server to compress answers (hence it might actually do it)
-K, --disable-keepalive disable keep-alive feature
--dns-cache cache DNS requests
-D, --dns-full-resolution enable full DNS resolution from the root servers
-d, --dns-server string specify an alternate DNS server for resolutions
-x, --extra-parameter extra changing parameter, add an extra changing parameter to the request to avoid being cached by reverse proxy
-F, --follow-redirects follow HTTP redirects (codes 3xx)
--head perform HTTP HEAD requests instead of GETs
-H, --header string add one or more header, in the form "name: value"
-h, --help help for http-ping
-1, --http1 use the HTTP/1 protocol
-2, --http2 use the HTTP/2 protocol
-3, --http3 use the HTTP/3 protocol
-k, --insecure allow insecure server connections when using SSL
-i, --interval duration define the wait time between each request (default 1s)
-4, --ipv4 force IPv4 resolution for dual-stacked sites
-6, --ipv6 force IPv6 resolution for dual-stacked sites
--keep-cookies keep received cookies between requests
--method string select a which HTTP method to be used (default "GET")
--no-server-error ignore server errors (5xx), do not handle them as "lost pings"
--parameter string add one or more parameters to the query, in the form name:value
-q, --quiet print less details
--referrer string define the referrer
-t, --throughput log the number of requests done per second
-T, --throughput-refresh duration sampling time for measuring throughput (default 5s)
--user-agent string define a custom user-agent (default "Http-Ping/(devel) (https://github.com/fever-ch/http-ping)")
-v, --verbose print more details
--version version for http-ping
-w, --wait duration define the time for a response before timing out (default 10s)
--workers int define the number of workers to be used (default 1)
Measure the latency with the Google Cloud Zurich region with 4 HTTP pings (-c 4):
$ http-ping https://europe-west6-5tkroniexa-oa.a.run.app/api/ping -c 4
HTTP-PING https://europe-west6-5tkroniexa-oa.a.run.app/api/ping GET
─→ server advertised HTTP/3 endpoint, using HTTP/3
1: HTTP/3.0, 216.239.32.53:443, code=200, size=13 bytes, time=28.4 ms
2: HTTP/3.0, 216.239.32.53:443, code=200, size=13 bytes, time=28.9 ms
3: HTTP/3.0, 216.239.32.53:443, code=200, size=13 bytes, time=29.5 ms
4: HTTP/3.0, 216.239.32.53:443, code=200, size=13 bytes, time=29.5 ms
--- https://europe-west6-5tkroniexa-oa.a.run.app/api/ping ping statistics ---
4 requests sent, 4 answers received, 0.0% loss
round-trip min/avg/max/stddev = 28.444/29.094/29.538/0.456 ms
Measure the latency with Google Cloud Zurich region with ten HTTP pings (-c 10), disabling socket reuse (-K), using
a HEAD request (-H), and in verbose mode (-v):
$ http-ping https://europe-west6-5tkroniexa-oa.a.run.app/api/ping -c 10 -K -H --http2 -v
HTTP-PING https://europe-west6-5tkroniexa-oa.a.run.app/api/ping HEAD
0: HTTP/2.0, 216.239.36.53:443, code=200, size=0 bytes, time=59.7 ms
proto=HTTP/2.0, socket reused=false, compressed=true
network i/o: bytes read=4713, bytes written=669
tls version=TLS-1.3
latency contributions:
59.7 ms request and response
├─ 39.1 ms connection setup
│ ├─ 3.9 ms DNS resolution
│ ├─ 9.0 ms TCP handshake
│ └─ 26.0 ms TLS handshake
├─ 0.9 ms request sending
├─ 17.3 ms wait
└─ 0.2 ms response ingestion
...
9: HTTP/2.0, 216.239.36.53:443, code=200, size=0 bytes, time=53.5 ms
proto=HTTP/2.0, socket reused=false, compressed=true
network i/o: bytes read=4713, bytes written=669
tls version=TLS-1.3
latency contributions:
53.5 ms request and response
├─ 35.2 ms connection setup
│ ├─ 3.4 ms DNS resolution
│ ├─ 8.7 ms TCP handshake
│ └─ 22.9 ms TLS handshake
├─ 0.1 ms request sending
├─ 18.0 ms wait
└─ 0.1 ms response ingestion
--- https://europe-west6-5tkroniexa-oa.a.run.app/api/ping ping statistics ---
10 requests sent, 10 answers received, 0.0% loss
round-trip min/avg/max/stddev = 52.900/56.048/59.653/2.089 ms
average latency contributions:
56.0 ms request and response
├─ 37.7 ms connection setup
│ ├─ 3.6 ms DNS resolution
│ ├─ 9.5 ms TCP handshake
│ └─ 24.5 ms TLS handshake
├─ 0.1 ms request sending
├─ 17.7 ms wait
└─ 0.2 ms response ingestion
note: the latency contribution tree only covers the main steps of the HTTP exchange, thus the sum doesn't fully match.
http-ping can also be used to measure the throughput given by an HTTP/S setup.
Beware, this measure generates, by design, a lot of traffic on the target, thus it might deteriorate the overall service during the time of measurement. This feature is more appropriate when you want to test your own servers.
Measure the maximum throughput to URL-TO-TEST by not waiting between request (-i 0s), by enabling throughput
measurement (-t) and, by allocating 64 workers (--workers 64)
> http-ping --workers 64 -i 0s -t URL-TO-TEST
HTTP-PING URL-TO-TEST GET
throughput: 1922.0 queries/sec, average latency: 32.6 ms
throughput: 2037.6 queries/sec, average latency: 31.4 ms
throughput: 1951.0 queries/sec, average latency: 32.8 ms
throughput: 2006.9 queries/sec, average latency: 31.8 ms
throughput: 1871.5 queries/sec, average latency: 34.1 ms
^C
--- URL-TO-TEST ping statistics ---
50470 requests sent, 50470 answers received, 0.0% loss
round-trip min/avg/max/stddev = 22.034/32.464/87.484/6.929 ms
throughput measures:
queries throughput min/avg/max/stdev = 1871.5/1957.8/2037.6/59.3 queries/sec The releases are providing packages for the following systems:
deb: Debian, Ubuntu, ...rpm: RedHat, CentOS, SuSE, ...apk: Alpine
n.b.: http-ping uses ca-certificates in order to authenticate the CA signature on the server certificates in
communications over HTTPS.
A third party repository, fever-ch/tap, provides up-to-date formula to
deploy http-ping easily on platforms supported by Homebrew.
To install using Brew, run the following command:
$ brew install fever-ch/tap/http-ping
The releases page also provides archives that contain the binary
executable file, tar.gz files for Linux/FreeBSD/MacOS and zip files for Windows.
$ docker run --rm feverch/http-ping
Note: images are published as feverch/http-ping (Central Docker registry) or ghcr.io/fever-ch/http-ping (Github
Container registry)
You can easily build http-ping, if golang is installed on your system.
$ go install fever.ch/http-ping@latest
HTTP/3 support is brought by a 3rd-party library QUIC-Go which APIs differs a bit
from the Go's HTTP (net/http) libraries that are used for HTTP/1 and HTTP/2.
Support for HTTP/3 came with version 1.2.0, and still needs to be considered as experimental.