| 1 |
NAME |
| 2 |
AnyEvent::FastPing - quickly ping a large number of hosts |
| 3 |
|
| 4 |
SYNOPSIS |
| 5 |
use AnyEvent::FastPing; |
| 6 |
|
| 7 |
DESCRIPTION |
| 8 |
This module was written for a single purpose only: sending ICMP ECHO |
| 9 |
REQUEST packets as quickly as possible to a large number of hosts |
| 10 |
(thousands to millions). |
| 11 |
|
| 12 |
It employs a separate thread and is fully event-driven (using AnyEvent), |
| 13 |
so you have to run an event model supported by AnyEvent to use this |
| 14 |
module. |
| 15 |
|
| 16 |
FUNCTIONS |
| 17 |
AnyEvent::FastPing::ipv4_supported |
| 18 |
Returns true iff IPv4 is supported in this module and on this |
| 19 |
system. |
| 20 |
|
| 21 |
AnyEvent::FastPing::ipv6_supported |
| 22 |
Returns true iff IPv6 is supported in this module and on this |
| 23 |
system. |
| 24 |
|
| 25 |
AnyEvent::FastPing::icmp4_pktsize |
| 26 |
Returns the number of octets per IPv4 ping packet (the whole IP |
| 27 |
packet including headers, excluding lower-level headers or trailers |
| 28 |
such as Ethernet). |
| 29 |
|
| 30 |
Can be used to calculate e.g. octets/s from rate ... |
| 31 |
|
| 32 |
my $octets_per_second = $packets_per_second * AnyEvent::FastPing::icmp4_pktsize; |
| 33 |
|
| 34 |
... or convert kilobit/second to packet rate ... |
| 35 |
|
| 36 |
my $packets_per_second = $kilobit_per_second |
| 37 |
* (1000 / 8 / AnyEvent::FastPing::icmp4_pktsize); |
| 38 |
|
| 39 |
etc. |
| 40 |
|
| 41 |
AnyEvent::FastPing::icmp6_pktsize |
| 42 |
Like AnyEvent::FastPing::icmp4_pktsize, but for IPv6. |
| 43 |
|
| 44 |
THE AnyEvent::FastPing CLASS |
| 45 |
The AnyEvent::FastPing class represents a single "pinger". A "pinger" |
| 46 |
comes with its own thread to send packets in the background, a |
| 47 |
rate-limit machinery and separate idle/receive callbacks. |
| 48 |
|
| 49 |
The recommended workflow (there are others) is this: 1. create a new |
| 50 |
AnyEvent::FastPing object 2. configure the address lists and ranges to |
| 51 |
ping, also configure an idle callback and optionally a receive callback |
| 52 |
3. "start" the pinger. |
| 53 |
|
| 54 |
When the pinger has finished pinging all the configured addresses it |
| 55 |
will call the idle callback. |
| 56 |
|
| 57 |
The pinging process works like this: every range has a minimum interval |
| 58 |
between sends, which is used to limit the rate at which hosts in that |
| 59 |
range are being pinged. Distinct ranges are independent of each other, |
| 60 |
which is why there is a per-pinger "global" minimum interval as well. |
| 61 |
|
| 62 |
The pinger sends pings as fats as possible, while both obeying the |
| 63 |
pinger rate limit as well as range limits. |
| 64 |
|
| 65 |
When a range is exhausted, it is removed. When all ranges are exhausted, |
| 66 |
the pinger waits another "max_rtt" seconds and then exits, causing the |
| 67 |
idle callback to trigger. |
| 68 |
|
| 69 |
Performance: On my 2 GHz Opteron system with a pretty average nvidia |
| 70 |
gigabit network card I can ping around 60k to 200k addresses per second, |
| 71 |
depending on routing decisions. |
| 72 |
|
| 73 |
Example: ping 10.0.0.1-10.0.0.15 with at most 100 packets/s, and |
| 74 |
11.0.0.1-11.0.255.255 with at most 1000 packets/s. Also ping the IPv6 |
| 75 |
loopback address 5 times as fast as possible. Do not, however, exceed |
| 76 |
1000 packets/s overall. Also dump each received reply. |
| 77 |
|
| 78 |
use AnyEvent::Socket; |
| 79 |
use AnyEvent::FastPing; |
| 80 |
|
| 81 |
my $done = AnyEvent->condvar; |
| 82 |
|
| 83 |
my $pinger = new AnyEvent::FastPing; |
| 84 |
|
| 85 |
$pinger->interval (1/1000); |
| 86 |
$pinger->max_rtt (0.1); # reasonably fast/reliable network |
| 87 |
|
| 88 |
$pinger->add_range (v10.0.0.1, v10.0.0.15, 1/100); |
| 89 |
$pinger->add_range (v11.0.0.1, v11.0.255.255, 1/1000); |
| 90 |
$pinger->add_hosts ([ (v0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.1) x 5 ]); |
| 91 |
|
| 92 |
$pinger->on_recv (sub { |
| 93 |
for (@{ $_[0] }) { |
| 94 |
printf "%s %g\n", (AnyEvent::Socket::format_address $_->[0]), $_->[1]; |
| 95 |
} |
| 96 |
}); |
| 97 |
|
| 98 |
$pinger->on_idle (sub { |
| 99 |
print "done\n"; |
| 100 |
undef $pinger; |
| 101 |
}); |
| 102 |
|
| 103 |
$pinger->start; |
| 104 |
$done->wait; |
| 105 |
|
| 106 |
METHODS |
| 107 |
$pinger = new AnyEvent::FastPing |
| 108 |
Creates a new pinger - right now there can be at most 65536 pingers |
| 109 |
in a process, although that limit might change to something |
| 110 |
drastically lower - you should be stingy with your pinger objects. |
| 111 |
|
| 112 |
$pinger->on_recv ($callback->([[$host, $rtt], ...])) |
| 113 |
Registers a callback to be called for ping replies. If no callback |
| 114 |
has been registered than ping replies will be ignored, otherwise |
| 115 |
this module calculates the round trip time, in seconds, for each |
| 116 |
reply and calls this callback. |
| 117 |
|
| 118 |
The callback receives a single argument, which is an array reference |
| 119 |
with an entry for each reply packet (the replies will be batched for |
| 120 |
efficiency). Each member in the array reference is again an array |
| 121 |
reference with exactly two members: the binary host address (4 |
| 122 |
octets for IPv4, 16 for IPv6) and the approximate round trip time, |
| 123 |
in seconds. |
| 124 |
|
| 125 |
The replies will be passed to the callback as soon as they arrive, |
| 126 |
and this callback can be called many times with batches of replies. |
| 127 |
|
| 128 |
The receive callback will be called whenever a suitable reply |
| 129 |
arrives, whether generated by this pinger or not, whether this |
| 130 |
pinger is started or not. The packets will have a unique 64 bit ID |
| 131 |
to distinguish them from other pinger objects and other generators, |
| 132 |
but this doesn't help against malicious replies. |
| 133 |
|
| 134 |
Note that very high packet rates can overwhelm your process, causing |
| 135 |
replies to be dropped (configure your kernel with long receive |
| 136 |
queues for raw sockets if this is a problem). |
| 137 |
|
| 138 |
Example: register a callback which simply dumps the received data. |
| 139 |
|
| 140 |
use AnyEvent::Socket; |
| 141 |
|
| 142 |
$pinger->on_recv (sub { |
| 143 |
for (@{ $_[0] }) { |
| 144 |
printf "%s %g\n", (AnyEvent::Socket::format_address $_->[0]), $_->[1]; |
| 145 |
} |
| 146 |
}); |
| 147 |
|
| 148 |
Example: a single ping reply with payload of 1 from "::1" gets |
| 149 |
passed like this: |
| 150 |
|
| 151 |
[ |
| 152 |
[ "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\1", 0.000280141830444336 ] |
| 153 |
] |
| 154 |
|
| 155 |
Example: ping replies for 127.0.0.1 and 127.0.0.2: |
| 156 |
|
| 157 |
[ |
| 158 |
[ "\177\0\0\1", 0.00015711784362793 ], |
| 159 |
[ "\177\0\0\2", 0.00090184211731 ] |
| 160 |
] |
| 161 |
|
| 162 |
$pinger->on_idle ($callback->()) |
| 163 |
Registers a callback to be called when the pinger becomes *idle*, |
| 164 |
that is, it has been started, has exhausted all ping ranges and |
| 165 |
waited for the "max_rtt" time. An idle pinger is also stopped, so |
| 166 |
the callback can instantly add new ranges, if it so desires. |
| 167 |
|
| 168 |
$pinger->interval ($seconds) |
| 169 |
Configures the minimum interval between packet sends for this pinger |
| 170 |
- the pinger will not send packets faster than this rate (or |
| 171 |
actually 1 / rate), even if individual ranges have a lower interval. |
| 172 |
|
| 173 |
A value of 0 selects the fastest possible speed (currently no faster |
| 174 |
than 1_000_000 packets/s). |
| 175 |
|
| 176 |
$pinger->max_rtt ($seconds) |
| 177 |
If your idle callback were called instantly after all ranges were |
| 178 |
exhausted and you destroyed the object inside (which is common), |
| 179 |
then there would be no chance to receive some replies, as there |
| 180 |
would be no time of the packet to travel over the network. |
| 181 |
|
| 182 |
This can be fixed by starting a timer in the idle callback, or more |
| 183 |
simply by selecting a suitable "max_rtt" value, which should be the |
| 184 |
maximum time you allow a ping packet to travel to its destination |
| 185 |
and back. |
| 186 |
|
| 187 |
The pinger thread automatically waits for this amount of time before |
| 188 |
becoming idle. |
| 189 |
|
| 190 |
The default is currently 0.5 seconds, which is usually plenty. |
| 191 |
|
| 192 |
$pinger->add_range ($lo, $hi[, $interval]) |
| 193 |
Ping the IPv4 (or IPv6, but see below) address range, starting at |
| 194 |
binary address $lo and ending at $hi (both $lo and $hi will be |
| 195 |
pinged), generating no more than one ping per $interval seconds (or |
| 196 |
as fast as possible if omitted). |
| 197 |
|
| 198 |
You can convert IP addresses from text to binary form by using |
| 199 |
"AnyEvent::Util::parse_address", "Socket::inet_aton", |
| 200 |
"Socket6::inet_pton" or any other method that you like :) |
| 201 |
|
| 202 |
The algorithm to select the next address is O(log n) on the number |
| 203 |
of ranges, so even a large number of ranges (many thousands) is |
| 204 |
manageable. |
| 205 |
|
| 206 |
No storage is allocated per address. |
| 207 |
|
| 208 |
Note that, while IPv6 addresses are currently supported, the |
| 209 |
usefulness of this option is extremely limited and might be gone in |
| 210 |
future versions - if you want to ping a number of IPv6 hosts, better |
| 211 |
specify them individually using the "add_hosts" method. |
| 212 |
|
| 213 |
$pinger->add_hosts ([$host...], $interval, $interleave) |
| 214 |
Similar to "add_range", but uses a list of single addresses instead. |
| 215 |
The list is specified as an array reference as first argument. Each |
| 216 |
entry in the array should be a binary host address, either IPv4 or |
| 217 |
IPv6. If all addresses are IPv4 addresses, then a compact IPv4-only |
| 218 |
format will be used to store the list internally. |
| 219 |
|
| 220 |
Minimum $interval is the same as for "add_range" and can be left |
| 221 |
out. |
| 222 |
|
| 223 |
$interlave specifies an increment between addresses: often address |
| 224 |
lists are generated in a way that results in clustering - first all |
| 225 |
addresses from one subnet, then from the next, and so on. To avoid |
| 226 |
this, you can specify an interleave factor. If it is 1 (the |
| 227 |
default), then every address is pinged in the order specified. If it |
| 228 |
is 2, then only every second address will be pinged in the first |
| 229 |
round, followed by a second round with the others. Higher factors |
| 230 |
will create $interleave runs of addresses spaced $interleave indices |
| 231 |
in the list. |
| 232 |
|
| 233 |
The special value 0 selects a (hopefully) suitable interleave factor |
| 234 |
automatically - currently 256 for lists with less than 65536 |
| 235 |
addresses, and the square root of the list length otherwise. |
| 236 |
|
| 237 |
$pinger->start |
| 238 |
Start the pinger, unless it is running already. While a pinger is |
| 239 |
running you must not modify the pinger. If you want to change a |
| 240 |
parameter, you have to "stop" the pinger first. |
| 241 |
|
| 242 |
The pinger will automatically stop when destroyed. |
| 243 |
|
| 244 |
$pinger->stop |
| 245 |
Stop the pinger, if it is running. A pinger can be stopped at any |
| 246 |
time, after which it's current state is preserved - starting it |
| 247 |
again will continue where it left off. |
| 248 |
|
| 249 |
AUTHOR |
| 250 |
Marc Lehmann <schmorp@schmorp.de> |
| 251 |
http://home.schmorp.de/ |
| 252 |
|
| 253 |
LICENSE |
| 254 |
This software is distributed under the GENERAL PUBLIC LICENSE, version 2 |
| 255 |
or any later version or, at your option, the Artistic License. |
| 256 |
|
