tc-netem - Man Page

tc qdisc ... dev DEVICE ] add netem Options

Options := [ LIMIT ] [ DELAY ] [ LOSS ] [ CORRUPT ] [ DUPLICATION ] [ REORDERING ] [ RATE ] [ SLOT ] [ SEED ]

LIMIT := limit packets

DELAY := delay TIME [ JITTER [ CORRELATION ]]]
      [ distribution { uniform | normal | pareto | paretonormal } ]

LOSS := loss { random PERCENT [ CORRELATION ]  |
              state p13 [ p31 [ p32 [ p23 [ p14]]]] |
              gemodel p [ r [ 1-h [ 1-k ]]] } [ ecn ]

CORRUPT := corrupt PERCENT [ CORRELATION ]]

DUPLICATION := duplicate PERCENT [ CORRELATION ]]

REORDERING := reorder PERCENT [ CORRELATION ] [ gap DISTANCE ]

RATE := rate RATE [ PACKETOVERHEAD [ CELLSIZE [ CELLOVERHEAD ]]]]

SLOT := slot { MIN_DELAY [ MAX_DELAY ] |
              distribution { uniform | normal | pareto | paretonormal | FILE } DELAY JITTER }
            [ packets PACKETS ] [ bytes BYTES ]

SEED := seed VALUE

The netem queue discipline provides Network Emulation functionality for testing protocols by emulating the properties of real-world networks.

The queue discipline provides one or more network impairments to packets such as: delay, loss, duplication, and packet corruption.

limit COUNT

Limits the maximum number of packets the qdisc may hold when doing delay.

delay

TIME [ JITTER [ CORRELATION ]]]
Delays the packets before sending. The optional parameters allow introducing a delay variation and a correlation. Delay and jitter values are expressed in milliseconds; Correlation is set by specifying a percent of how much the previous delay will impact the current random value.

distribution TYPE

Specifies a pattern for delay distribution.

uniform

Use an equally weighted distribution of packet delays.

normal

Use a Gaussian distribution of delays. Sometimes called a Bell Curve.

pareto

Use a Pareto distribution of packet delays. This is useful to emulate long-tail distributions.

paretonormal

This is a mix of pareto and normal distribution which has properties of both Bell curve and long tail.

loss MODEL

Drop packets based on a loss model. MODEL can be one of

random PERCENT

Each packet loss is independent.

state P13 [ P31 [ P32 [ P23 P14 ]]]

Use a 4-state Markov chain to describe packet loss.
P13 is the packet loss. Optional parameters extend the model to 2-state P31, 3-state P23, P32 and 4-state P14.

The Markov chain states are:

1. good packet reception (no loss).
2. good reception within a burst.
3. burst losses.
4. independent losses.

gemodel PERCENT [ R [ 1-H [ 1-K ]]]

Use a Gilbert-Elliot (burst loss) model based on:

PERCENT

probability of starting bad (lossy) state.

R

probability of exiting bad state.

1-H

loss probability in bad state.

1-K

loss probability in good state.

ecn

Use Explicit Congestion Notification (ECN) to mark packets instead of dropping them. A loss model has to be used for this to be enabled.

corrupt PERCENT

modifies the contents of the packet at a random position based on PERCENT.

duplicate PERCENT

creates a copy of the packet before queuing.

reorder PERCENT

modifies the order of packet in the queue.

gap DISTANCE

sends some packets immediately. The first packets (DISTANCE - 1) are delayed and the next packet is sent immediately.

rate RATE [ PACKETOVERHEAD [ CELLSIZE  [ CELLOVERHEAD ]]]

Delays packets based on packet size to emulate a fixed link speed. Optional parameters:

PACKETOVERHEAD

Specify a per packet overhead in bytes. Used to simulate additional link layer headers. A negative value can be used to simlate when the Ethernet header is stripped (e.g. -14) or header compression is used.

CELLSIZE

simulate link layer schemes like ATM.

CELLOVERHEAD

specify per cell overhead.

Rate throttling impacted by several factors including the kernel clock granularity. This will show up in an artificial packet compression (bursts).

slot MIN_DELAY [  MAX_DELAY  ]

allows emulating slotted networks. Defer delivering accumulated packets to within a slot. Each available slot is configured with a minimum delay to acquire, and an optional maximum delay.

slot distribution

allows configuring based on distribution similar to distribution option for packet delays.

These slot options can provide a crude approximation of bursty MACs such as DOCSIS, WiFi, and LTE.

Slot emulation is limited by several factors: the kernel clock granularity, as with a rate, and attempts to deliver many packets within a slot will be smeared by the timer resolution, and by the underlying native bandwidth also.

It is possible to combine slotting with a rate, in which case complex behaviors where either the rate, or the slot limits on bytes or packets per slot, govern the actual delivered rate.

seed VALUE

Specifies a seed to guide and reproduce the randomly generated loss or corruption events.

Netem is limited by the timer granularity in the kernel. Rate and delay maybe impacted by clock interrupts.

Mixing forms of reordering may lead to unexpected results. For any method of reordering to work, some delay is necessary. If the delay is less than the inter-packet arrival time then no reordering will be seen. Due to mechanisms like TSQ (TCP Small Queues), for TCP performance test results to be realistic netem must be placed on the ingress of the receiver host.

Combining netem with other qdisc is possible but may not always work because netem use skb control block to set delays.

# tc qdisc add dev eth0 root netem delay 100ms

# tc qdisc change dev eth0 root netem delay 100ms 10ms 25%

# tc qdisc change dev eth0 root netem delay 100ms 20ms distribution normal

# tc qdisc change dev eth0 root netem loss 0.1%

# tc qdisc change dev eth0 root netem duplicate 1%

# tc qdisc change dev eth0 root netem loss 0.3% 25%

There are two different ways to specify reordering. The gap method uses a fixed sequence and reorders every Nth packet.

# tc qdisc change dev eth0 root netem gap 5 delay 10ms

The reorder form uses a percentage of the packets to get misordered.

# tc qdisc change dev eth0 root netem delay 10ms reorder 25% 50%

In this example, 25% of packets (with a correlation of 50%) will get sent immediately, others will be delayed by 10ms.

Packets will also get reordered if jitter is large enough.

# tc qdisc change dev eth0 root netem delay 100ms 75ms

If you don't want this behavior then replace the internal queue discipline tfifo with a simple FIFO queue discipline.

# tc qdisc add dev eth0 root handle 1: netem delay 10ms 100ms
# tc qdisc add dev eth0 parent 1:1 pfifo limit 1000

Example of using rate control and cells size.

# tc qdisc add dev eth0 root netem rate 5kbit 20 100 5

It is possible to selectively apply impairment using traffic classification.

# tc qdisc add dev eth0 root handle 1: prio
# tc qdisc add dev eth0 parent 1:3 handle 30:    tbf rate 20kbit buffer 1600 limit  3000
# tc qdisc add dev eth0 parent 30:1 handle 31:    netem delay 200ms 10ms distribution normal
# tc filter add dev eth0 protocol ip parent 1:0 prio 3 u32    match ip dst 65.172.181.4/32 flowid 1:3

1. Hemminger S. , "Network Emulation with NetEm", Open Source Development Lab, April 2005 http://devresources.linux-foundation.org/shemminger/netem/LCA2005_paper.pdf
2. Salsano S., Ludovici F., Ordine A., "Definition of a general and intuitive loss model for packet networks and its implementation in the Netem module in the Linux kernel", available at http://netgroup.uniroma2.it/NetemCLG

tc(8)

Netem was written by Stephen Hemminger at Linux foundation and was inspired by NISTnet.

Original manpage was created by Fabio Ludovici <fabio.ludovici at yahoo dot it> and Hagen Paul Pfeifer <hagen@jauu.net>.

ovs-vswitchd.conf.db(5).