1TC(8)                                Linux                               TC(8)
2
3
4

NAME

6       drr - deficit round robin scheduler
7

SYNOPSIS

9       tc qdisc ... add drr [ quantum bytes ]
10
11

DESCRIPTION

13       The Deficit Round Robin Scheduler is a classful queuing discipline as a
14       more flexible replacement for Stochastic Fairness Queuing.
15
16       Unlike SFQ, there are no built-in queues -- you need to add classes and
17       then  set up filters to classify packets accordingly.  This can be use‐
18       ful e.g. for using RED qdiscs with different  settings  for  particular
19       traffic. There is no default class -- if a packet cannot be classified,
20       it is dropped.
21
22

ALGORITHM

24       Each class is assigned a deficit counter, initialized to quantum.
25
26       DRR maintains an (internal) ''active'' list of classes whose qdiscs are
27       non-empty.  This  list is used for dequeuing. A packet is dequeued from
28       the class at the head of the list if the  packet  size  is  smaller  or
29       equal  to  the  deficit  counter.  If  the  counter is too small, it is
30       increased by quantum and the scheduler moves on to the  next  class  in
31       the active list.
32
33
34

PARAMETERS

36       quantum
37              Amount  of  bytes a flow is allowed to dequeue before the sched‐
38              uler moves to the next class. Defaults to the MTU of the  inter‐
39              face. The minimum value is 1.
40
41

EXAMPLE & USAGE

43       To attach to device eth0, using the interface MTU as its quantum:
44
45       # tc qdisc add dev eth0 handle 1 root drr
46
47       Adding two classes:
48
49       # tc class add dev eth0 parent 1: classid 1:1 drr
50       # tc class add dev eth0 parent 1: classid 1:2 drr
51
52       You also need to add at least one filter to classify packets.
53
54       # tc filter add dev eth0 protocol .. classid 1:1
55
56       Like  SFQ,  DRR  is  only useful when it owns the queue -- it is a pure
57       scheduler and does not  delay  packets.  Attaching  non-work-conserving
58       qdiscs like tbf to it does not make sense -- other qdiscs in the active
59       list will also become inactive until the  dequeue  operation  succeeds.
60       Embed  DRR  within another qdisc like HTB or HFSC to ensure it owns the
61       queue.
62
63       You can mimic SFQ behavior by assigning packets to the attached classes
64       using the flow filter:
65
66       tc qdisc add dev .. drr
67
68       for i in .. 1024;do
69            tc class add dev .. classid $handle:$(print %x $i)
70            tc qdisc add dev .. fifo limit 16
71       done
72
73       tc   filter   add   ..   protocol   ip   ..   $handle  flow  hash  keys
74       src,dst,proto,proto-src,proto-dst divisor 1024 perturb 10
75
76
77

SOURCE

79       o      M. Shreedhar and George Varghese "Efficient Fair  Queuing  using
80              Deficit Round Robin", Proc. SIGCOMM 95.
81
82

NOTES

84       This  implementation  does  not  drop packets from the longest queue on
85       overrun, as limits are handled by the individual child qdiscs.
86
87

SEE ALSO

89       tc(8), tc-htb(8), tc-sfq(8)
90
91

AUTHOR

93       sched_drr was written by Patrick McHardy.
94
95
96
97iproute2                         January 2010                            TC(8)