perfdhcp - DHCP benchmarking tool¶
perfdhcp [-1] [-4 | -6] [-A encapsulation-level] [-b base] [-B] [-c] [-C separator] [-d drop-time] [-D max-drop] [-e lease-type] [-E time-offset] [-f renew-rate] [-F release-rate] [-g thread-mode] [-h] [-i] [-I ip-offset] [-J giaddr-list-file] [-l local-address|interface] [-L local-port] [-M mac-list-file] [-n num-request] [-N remote-port] [-O random-offset] [-o code,hexstring] [-p test-period] [-P preload] [-r rate] [-R num-clients] [-s seed] [-S srvid-offset] [–scenario name] [-t report] [-T template-file] [-u] [-v] [-W exit-wait-time] [-w script_name] [-x diagnostic-selector] [-X xid-offset] [server]
perfdhcp is a DHCP benchmarking tool. It provides a way of measuring
the performance of DHCP servers by generating large amounts of traffic
from simulated multiple clients. It is able to test both IPv4 and IPv6
servers, and provides statistics concerning response times and the
number of requests that are dropped.
The tool supports scenarios, which offer certain behaviours.
By default (basic scenario) tests are run using the full four-packet exchange sequence
(DORA for DHCPv4, SARR for DHCPv6). An option is provided to run tests
using the initial two-packet exchange (DO and SA) instead. It is also
possible to configure
perfdhcp to send DHCPv6 RENEW and RELEASE messages
at a specified rate in parallel with the DHCPv6 four-way exchanges. By
default, if there is no response received with 1 second, a response is
considered lost and perfdhcp continues with other transactions.
Second scenario is called avalanche, which is selected by
It first sends as many Discovery or Solicit messages as request in -R option then
a retransmission (with exponential back off mechanism) is used for each simulated client until all requests are
answered. It will generate report when all clients get their addresses or when
it will be manually stopped. This scenario attempts to replicate a
case where the server is not able to handle the traffic swiftly
enough. Real clients will assume the packet or the response was lost
and will retransmit, further increasing DHCP traffic. This is
sometimes called avalanche effect, thus the scenario name.
-p is ignored in avalanche scenario.
When running a performance test,
perfdhcp will exchange packets with
the server under test as fast as possible unless the
-r parameter is used to
limit the request rate. The length of the test can be limited by setting
a threshold on any or all of the number of requests made by
perfdhcp, the elapsed time, or the number of requests dropped by the
To allow the contents of packets sent to the server to be customized,
perfdhcp allows the specification of template files that determine
the contents of the packets. For example, the customized packet may
contain a DHCPv6 ORO to request a set of options to be returned by the
server, or it may contain the Client FQDN option to request that the server
perform DNS updates. This may be used to discover performance
bottlenecks for different server configurations (e.g. DDNS enabled or
Up to two template files can be specified on the command line, each file representing the contents of a particular type of packet, the type being determined by the test being carried out. For example, if testing DHCPv6:
- With no template files specified on the command line,
perfdhcpwill generate both SOLICIT and REQUEST packets.
- With one template file specified, that file will be used as the
pattern for SOLICIT packets:
perfdhcpwill generate the REQUEST packets.
- With two template files given on the command line, the first will be used as the pattern for SOLICIT packets, the second as the pattern for REQUEST packets.
(Similar determination applies to DHCPv4’s DISCOVER and REQUEST packets.)
The template file holds the DHCP packet represented as a stream of ASCII hexadecimal digits and it excludes any IP/UDP stack headers. The template file must not contain any characters other than hexadecimal digits and spaces. Spaces are discarded when the template file is parsed; in the file, ‘12B4’ is the same as ‘12 B4’ which is the same as ‘1 2 B 4’.
The template files should be used in conjunction with the command-line
parameters which specify offsets of the data fields being modified in
outbound packets. For example, the
-E time-offset switch specifies
the offset of the DHCPv6 Elapsed Time option in the packet template.
If the offset is specified,
perfdhcp will inject the current elapsed-time
value into this field before sending the packet to the server.
In many scenarios,
perfdhcp needs to simulate multiple clients,
each having a unique client identifier. Since packets for each client are
generated from the same template file, it is necessary to randomize the
client identifier (or HW address in DHCPv4) in the packet created from
-O random-offset option allows specification of the offset in
the template where randomization should be performed. It is important to
note that this offset points to the end (not the beginning) of the
client identifier (or HW address field). The number of bytes being
randomized depends on the number of simulated clients. If the number of
simulated clients is between 1 and 255, only one byte (to which the
randomization offset points) will be randomized. If the number of
simulated clients is between 256 and 65535, two bytes will be
randomized. Note that the last two bytes of the client identifier will be
randomized in this case: the byte which the randomization offset parameter
points to, and the one which precedes it (random-offset - 1). If the
number of simulated clients exceeds 65535, three bytes will be
randomized, and so on.
Perfdhcp can now simulate traffic from multiple subnets by enabling option -J and passing path to file that contains v4 addresses that will be used as giaddr in generated messages. That enable testing of vast numbers of Kea shared networks. Kea should be started with KEA_TEST_SEND_RESPONSES_TO_SOURCE environment variable to force Kea to send generated messages to source address of incoming packet. Feature is not available in kea-dhcp6.
Templates may currently be used to generate packets being sent to the server in 4-way exchanges, i.e. SOLICIT, REQUEST (DHCPv6) and DISCOVER, REQUEST (DHCPv4). They cannot be used when RENEW or RELEASE packets are being sent.
- Takes the server-ID option from the first received message.
- Establishes DHCPv4 operation; this is the default. It is incompatible with the
- Establishes DHCPv6 operation. This is incompatible with the
- Indicates the base MAC or DUID used to simulate different clients. The basetype
may be “mac” or “duid”. (The keyword “ether” may alternatively used
for MAC.) The
-boption can be specified multiple times. The MAC address must consist of six octets separated by single (:) or double (::) colons, for example: mac=00:0c:01:02:03:04. The DUID value is a hexadecimal string; it must be at least six octets long and not longer than 64 bytes, and the length must be less than 128 hexadecimal digits, for example: duid=0101010101010101010110111F14.
- Specifies the time after which a request is treated as having been lost. The value is given in seconds and may contain a fractional component. The default is 1 second.
Specifies the type of lease being requested from the server. It may be one of the following:
- Only regular addresses (v4 or v6) will be requested.
- Only IPv6 prefixes will be requested.
- Both IPv6 addresses and prefixes will be requested.
-e address-and-prefixforms may not be used with the
- Specifies the rate at which DHCPv4 or DHCPv6 renew requests are sent to a server.
This value is only valid when used in conjunction with the exchange
rate (given by
-r rate). Furthermore, the sum of this value and the release-rate (given by
-F rate) must be equal to or less than the exchange rate.
- Allows selection of thread-mode, which can be either ‘single’ or ‘multi’. In multi-thread mode packets are received in a separate thread, which allows better utilisation of CPUs. In a single-CPU system it is better to run in one thread to avoid threads blocking each other. If more than one CPU is present in the system, multi-thread mode is the default; otherwise single-thread is the default.
- Prints help and exits.
Performs only the initial part of the exchange: DISCOVER-OFFER if
-4is selected, SOLICIT-ADVERTISE if
-iis incompatible with the following options:
-F. In addition, it cannot be used with multiple instances of
- For DHCPv4 operation, specifies the local hostname/address to use when communicating with the server. By default, the interface address through which traffic would normally be routed to the server is used. For DHCPv6 operation, specifies the name of the network interface through which exchanges are initiated.
- Specifies the local port to use. This must be zero or a positive
integer up to 65535. A value of 0 (the default) allows
perfdhcpto choose its own port.
- Specifies a text file containing a list of MAC addresses, one per line. If provided, a MAC address will be chosen randomly from this list for every new exchange. In DHCPv6, MAC addresses are used to generate DUID-LLs. This parameter must not be used in conjunction with the -b parameter.
- Specifies the remote port to use. This must be zero or a positive
integer up to 65535. A value of 0 (the default) allows
perfdhcpto choose the standard service port.
perfdhcpto insert the specified extra option (or options if used several times) into packets being transmitted. The code specifies the option code and the hexstring is a hexadecimal string that defines the content of the option. Care should be taken as
perfdhcpdoes not offer any kind of logic behind those options; they are simply inserted into packets and sent as is. Be careful not to duplicate options that are already inserted. For example, to insert client class identifier (option code 60) with a string ‘docsis’, use -o 60,646f63736973. The
-omay be used multiple times. It is necessary to specify the protocol family (either
-6) before using
- Initiates preload exchanges back-to-back at startup. Must be 0 (the default) or a positive integer.
- Initiates the rate of DORA/SARR (or if
-iis given, DO/SA) exchanges per second. A periodic report is generated showing the number of exchanges which were not completed, as well as the average response latency. The program continues until interrupted, at which point a final report is generated.
- Specifies how many different clients are used. With a value of 1 (the default), all requests seem to come from the same client. Must be a positive number.
- Specifies the seed for randomization, making runs of
perfdhcprepeatable. This must be 0 or a positive integer. The value 0 means that a seed is not used; this is the default.
- Specifies type of the scenario, can be basic (default) or avalanche.
- Specifies a file containing the template to use as a stream of hexadecimal digits. This may be specified up to two times and controls the contents of the packets sent (see the “Templates” section above).
- Enable checking address uniqueness. Lease valid lifetime should not be shorter than test duration and clients should not request address more than once without releasing it first.
- Prints the version of this program.
- Specifies the exit-wait-time parameter, which causes
perfdhcpto wait for exit-wait-time after an exit condition has been met, to receive all packets without sending any new packets. Expressed in microseconds. If not specified, 0 is used (i.e. exit immediately after exit conditions are met).
- Specifies the name of the script to be run before/after
perfdhcp. When called, the script is passed a single parameter, either “start” or “stop”, indicating whether it is being called before or after
Includes extended diagnostics in the output. This is a string of single keywords specifying the operations for which verbose output is desired. The selector key letters are:
- Prints the decoded command line arguments.
- Prints the exit reason.
- Prints the rate processing details.
- Prints the first server-ID.
- When finished, prints timers of all successful exchanges.
- When finished, prints templates.
- Time in seconds after which perfdhcp will start simulating the client waiting longer for server responses. This increase the secs field in DHCPv4 and sends increased values in Elapsed option in DHCPv6. Must be used with ‘-Y’.
- Period of time in seconds in which perfdhcp will be simulating the client waiting longer for server responses. This increase the secs field in DHCPv4 and sends increased values in Elapsed option in DHCPv6. Must be used with ‘-y’.
The following options only apply for DHCPv4 (i.e. when
-4 is given).
- Forces broadcast handling.
- Text file that include multiple addresses. If provided perfdhcp will choose randomly one of addresses for each exchange. This is used to generate traffic from multiple subnets. Designed to test shared-networks in kea-dhcp4. Kea should be started with KEA_TEST_SEND_RESPONSES_TO_SOURCE=ENABLE env variable otherwise perfdhcp will not be able to receive responses.
The following options only apply for DHCPv6 (i.e. when
-6 is given).
- Adds a rapid-commit option (exchanges will be SOLICIT-ADVERTISE).
- Specifies the rate at which IPv6 RELEASE requests are sent to a server. This value
is only valid when used in conjunction with the exchange rate (given
-r rate). Furthermore, the sum of this value and the renew-rate (given by
-f rate) must be equal to or less than the exchange rate value.
- Specifies that relayed traffic must be generated. The argument specifies the level of encapsulation, i.e. how many relay agents are simulated. Currently the only supported encapsulation-level value is 1, which means that the generated traffic is equivalent to the amount of traffic passing through a single relay agent.
Options Controlling a Test¶
- Aborts the test immediately if max-drop requests have been dropped.
-D 0to abort if even a single request has been dropped. max-drop must be a positive integer. If max-drop includes the suffix ‘%’, it specifies a maximum percentage of requests that may be dropped before abort. In this case, testing of the threshold begins after 10 requests have been expected to be received.
- Initiates num-request transactions. No report is generated until all transactions have been initiated/waited-for, after which a report is generated and the program terminates.
- Sends requests for test-period, which is specified in the same manner
-d. This can be used as an alternative to
-nor both options can be given, in which case the testing is completed when either limit is reached.
- Sets the delay (in seconds) between two successive reports.
- Output reduced, an argument is a separator for periodic (-t) reports generated in easy parsable mode. Data output won’t be changed, remain identical as in -t option.
- Indicates the server to test, specified as an IP address. In the DHCPv6 case, the
special name ‘all’ can be used to refer to
All_DHCP_Relay_Agents_and_Servers (the multicast address FF02::1:2),
or the special name ‘servers’ to refer to All_DHCP_Servers (the
multicast address FF05::1:3). The server is mandatory except where
-loption is given to specify an interface, in which case it defaults to ‘all’.
perfdhcp can report the following errors in the packet exchange:
- A message was received that was too short.
- A message was received which does not match one sent to the server (i.e. it is a duplicate message, a message that has arrived after an excessive delay, or one that is just not recognized).
- Local system limits have been reached when sending a message.
perfdhcp can exit with one of the following status codes:
- General error.
- Error in command-line arguments.
- No general failures in operation, but one or more exchanges were unsuccessful.
Simulate regular DHCPv4 traffic: 100 DHCPv4 devices (-R 100), 10 packets per second (-r 10), show the query/response rate details (-xi), the report should be shown every 2 seconds (-t 2), send the packets to the IP 192.0.2.1:
sudo perfdhcp -xi -t 2 -r 10 -R 100 192.0.2.1
Here’s a similar case, but for DHCPv6. Note that DHCPv6 protocol uses link-local addresses, so you need to specify the interface (eth0 in this example) to send the traffic. ‘all’ is a convenience alias for All_DHCP_Relay_Agents_and_Servers (the multicast address FF02::1:2). Alternatively, you can use ‘servers’ alias to refer to All_DHCP_Servers (the multicast address FF05::1:3), or skip it all together and the default value (all) will be used.
sudo perfdhcp -6 -xi -t 1 -r 1 -R 10 -l eth0 all
The following examples simulate normal DHCPv4 and DHCPv6 traffic that after 3 seconds starts pretending to not receive any responses from the server for 10 seconds. DHCPv4 protocol signals this by increased secs field and DHCPv6 uses elapsed option for that. In real networks this indicates that the clients are not getting responses in a timely matter. This can be used to simulate some HA scenarios, as Kea uses secs field and elapsed option value as one of the indicators that the HA partner is not responding. When enabled with -y and -Y, the secs and elapsed time value increased steadily.
sudo perfdhcp -xi -t 1 -r 1 -y 10 -Y 3 192.0.2.1
sudo perfdhcp -6 -xi -t 1 -r 1 -y 10 -Y 3 2001:db8::1
Mailing Lists and Support¶
There are two public mailing lists available for the Kea project. kea-users (kea-users at lists.isc.org) is intended for Kea users, while kea-dev (kea-dev at lists.isc.org) is intended for Kea developers, prospective contributors, and other advanced users. Both lists are available at https://lists.isc.org. The community provides best-effort support on both of those lists.
ISC provides professional support for Kea services. See https://www.isc.org/kea/ for details.
perfdhcp tool was initially coded in October 2011 by John
DuBois, Francis Dupont, and Marcin Siodelski of ISC. Kea 1.0.0, which
perfdhcp, was released in December 2015.
kea-dhcp4(8), kea-dhcp6(8), kea-dhcp-ddns(8), kea-ctrl-agent(8), kea-admin(8), kea-netconf(8), keactrl(8), kea-lfc(8), Kea Administrator Reference Manual.