1FQ(8) Linux FQ(8)
2
6 FQ - Fair Queue traffic policing
7
9 tc qdisc ... fq [ limit PACKETS ] [ flow_limit PACKETS ] [ quantum
10 BYTES ] [ initial_quantum BYTES ] [ maxrate RATE ] [ buckets NUMBER ]
11 [ pacing | nopacing ]
12
15 FQ (Fair Queue) is a classless packet scheduler meant to be mostly used
16 for locally generated traffic. It is designed to achieve per flow pac‐
17 ing. FQ does flow separation, and is able to respect pacing require‐
18 ments set by TCP stack. All packets belonging to a socket are consid‐
19 ered as a 'flow'. For non local packets (router workload), packet
20 rxhash is used as fallback.
21 An application can specify a maximum pacing rate using the SO_MAX_PAC‐
23 ING_RATE setsockopt call. This packet scheduler adds delay between
24 packets to respect rate limitation set by TCP stack.
25 Dequeueing happens in a round-robin fashion. A special FIFO queue is
27 reserved for high priority packets ( TC_PRIO_CONTROL priority), such
28 packets are always dequeued first.
29 FQ is non-work-conserving.
31 TCP pacing is good for flows having idle times, as the congestion win‐
33 dow permits TCP stack to queue a possibly large number of packets.
34 This removes the 'slow start after idle' choice, badly hitting large
35 BDP flows and applications delivering chunks of data such as video
36 streams.
37
40 limit
41 Hard limit on the real queue size. When this limit is reached, new
42 packets are dropped. If the value is lowered, packets are dropped so
43 that the new limit is met. Default is 10000 packets.
44 flow_limit
46 Hard limit on the maximum number of packets queued per flow. Default
47 value is 100.
48 quantum
50 The credit per dequeue RR round, i.e. the amount of bytes a flow is
51 allowed to dequeue at once. A larger value means a longer time period
52 before the next flow will be served. Default is 2 * interface MTU
53 bytes.
54 initial_quantum
56 The initial sending rate credit, i.e. the amount of bytes a new flow is
57 allowed to dequeue initially. This is specifically meant to allow
58 using IW10 without added delay. Default is 10 * interface MTU, i.e.
59 15140 for 'standard' ethernet.
60 maxrate
62 Maximum sending rate of a flow. Default is unlimited. Application
63 specific setting via SO_MAX_PACING_RATE is ignored only if it is larger
64 than this value.
65 buckets
67 The size of the hash table used for flow lookups. Each bucket is
68 assigned a red-black tree for efficient collision sorting. Default:
69 1024.
70 [no]pacing
72 Enable or disable flow pacing. Default is enabled.
73
75 #tc qdisc add dev eth0 root fq
76 #tc -s -d qdisc
77 qdisc fq 8003: dev eth0 root refcnt 2 limit 10000p flow_limit 100p
78 buckets 1024 quantum 3028 initial_quantum 15140
79 Sent 503727981 bytes 1146972 pkt (dropped 0, overlimits 0 requeues
80 54452)
81 backlog 0b 0p requeues 54452
82 1289 flows (1289 inactive, 0 throttled)
83 0 gc, 31 highprio, 27411 throttled
84
86 tc(8), socket(7)
87
89 FQ was written by Eric Dumazet.
90iproute2 10 Sept 2015 FQ(8)