1TC(8) Linux TC(8)
2
6 tbf - Token Bucket Filter
7
9 tc qdisc ... tbf rate rate burst bytes/cell ( latency ms | limit bytes
10 ) [ mpu bytes [ peakrate rate mtu bytes/cell ] ]
11 burst is also known as buffer and maxburst. mtu is also known as min‐
13 burst.
14
16 The Token Bucket Filter is a classful queueing discipline available for
17 traffic control with the tc(8) command.
18 TBF is a pure shaper and never schedules traffic. It is non-work-con‐
20 serving and may throttle itself, although packets are available, to
21 ensure that the configured rate is not exceeded. It is able to shape
22 up to 1mbit/s of normal traffic with ideal minimal burstiness, sending
23 out data exactly at the configured rates.
24 Much higher rates are possible but at the cost of losing the minimal
26 burstiness. In that case, data is on average dequeued at the configured
27 rate but may be sent much faster at millisecond timescales. Because of
28 further queues living in network adaptors, this is often not a problem.
29
32 As the name implies, traffic is filtered based on the expenditure of
33 tokens. Tokens roughly correspond to bytes, with the additional con‐
34 straint that each packet consumes some tokens, no matter how small it
35 is. This reflects the fact that even a zero-sized packet occupies the
36 link for some time.
37 On creation, the TBF is stocked with tokens which correspond to the
39 amount of traffic that can be burst in one go. Tokens arrive at a
40 steady rate, until the bucket is full.
41 If no tokens are available, packets are queued, up to a configured
43 limit. The TBF now calculates the token deficit, and throttles until
44 the first packet in the queue can be sent.
45 If it is not acceptable to burst out packets at maximum speed, a
47 peakrate can be configured to limit the speed at which the bucket emp‐
48 ties. This peakrate is implemented as a second TBF with a very small
49 bucket, so that it doesn't burst.
50 To achieve perfection, the second bucket may contain only a single
52 packet, which leads to the earlier mentioned 1mbit/s limit.
53 This limit is caused by the fact that the kernel can only throttle for
55 at minimum 1 'jiffy', which depends on HZ as 1/HZ. For perfect shaping,
56 only a single packet can get sent per jiffy - for HZ=100, this means
57 100 packets of on average 1000 bytes each, which roughly corresponds to
58 1mbit/s.
59
62 See tc(8) for how to specify the units of these values.
63 limit or latency
65 Limit is the number of bytes that can be queued waiting for
66 tokens to become available. You can also specify this the other
67 way around by setting the latency parameter, which specifies the
68 maximum amount of time a packet can sit in the TBF. The latter
69 calculation takes into account the size of the bucket, the rate
70 and possibly the peakrate (if set). These two parameters are
71 mutually exclusive.
72 burst Also known as buffer or maxburst. Size of the bucket, in bytes.
74 This is the maximum amount of bytes that tokens can be available
75 for instantaneously. In general, larger shaping rates require a
76 larger buffer. For 10mbit/s on Intel, you need at least 10kbyte
77 buffer if you want to reach your configured rate!
78 If your buffer is too small, packets may be dropped because more
80 tokens arrive per timer tick than fit in your bucket. The mini‐
81 mum buffer size can be calculated by dividing the rate by HZ.
82 Token usage calculations are performed using a table which by
84 default has a resolution of 8 packets. This resolution can be
85 changed by specifying the cell size with the burst. For example,
86 to specify a 6000 byte buffer with a 16 byte cell size, set a
87 burst of 6000/16. You will probably never have to set this. Must
88 be an integral power of 2.
89 mpu A zero-sized packet does not use zero bandwidth. For ethernet,
91 no packet uses less than 64 bytes. The Minimum Packet Unit
92 determines the minimal token usage (specified in bytes) for a
93 packet. Defaults to zero.
94 rate The speed knob. See remarks above about limits! See tc(8) for
96 units.
97 Furthermore, if a peakrate is desired, the following parameters are
99 available:
100 peakrate
103 Maximum depletion rate of the bucket. The peakrate does not need
104 to be set, it is only necessary if perfect millisecond timescale
105 shaping is required.
106 mtu/minburst
109 Specifies the size of the peakrate bucket. For perfect accuracy,
110 should be set to the MTU of the interface. If a peakrate is
111 needed, but some burstiness is acceptable, this size can be
112 raised. A 3000 byte minburst allows around 3mbit/s of peakrate,
113 given 1000 byte packets.
114 Like the regular burstsize you can also specify a cell size.
116
118 To attach a TBF with a sustained maximum rate of 0.5mbit/s, a peakrate
119 of 1.0mbit/s, a 5kilobyte buffer, with a pre-bucket queue size limit
120 calculated so the TBF causes at most 70ms of latency, with perfect
121 peakrate behaviour, issue:
122 # tc qdisc add dev eth0 handle 10: root tbf rate 0.5mbit \
124 burst 5kb latency 70ms peakrate 1mbit \
125 minburst 1540
126 To attach an inner qdisc, for example sfq, issue:
128 # tc qdisc add dev eth0 parent 10:1 handle 100: sfq
130 Without inner qdisc TBF queue acts as bfifo. If the inner qdisc is
132 changed the limit/latency is not effective anymore.
133
135 tc(8)
136
139 Alexey N. Kuznetsov, <kuznet@ms2.inr.ac.ru>. This manpage maintained by
140 bert hubert <ahu@ds9a.nl>
141iproute2 13 December 2001 TC(8)