WB 15.15

3 Group table

Groups represent sets of actions for flooding as well as more complex forwarding semantics (e.g.

multipath, fast reroute, and link aggregation). As a general layer of indirection, groups also enable

multiple flow entries to forward to a single identifier (e.g. IP forwarding to a common next hop).

This abstraction allows common output actions across flow entries to be changed efficiently.

The group table contains group entries; each group entry contains a list of action buckets with

specific semantics dependent on group type. The actions in one or more action buckets are applied

to packets sent to the group. There are 4 types of groups:

1. All

All the action buckets in the group should be executed when a packet hits the group table.

2. Select

Execute any one action bucket in the group. The switch implementation uses round-robin to

select the action bucket to be executed. Openflow specification defines a weight mechanism

to do load sharing. However, this is not supported in the switch implementation. The weight

MUST be given as 1. For all the other groups, weight MUST be specified as 0.

3. Indirect

Execute the one defined bucket in this group. This group supports only a single bucket.

4. Fast failover

Execute the first live bucket. The buckets are evaluated for liveness in the order defined by the

group.

For the implementation of groups, the following is important to note:

• Group table is supported only in software. Hence, group cannot be referenced directly from

a hardware flow entry.

• The number of groups per OpenFlow instance is capped to 32.

• The total number of groups in the switch is capped to 1024.

• Chaining of groups not supported. As a consequence, watch_group is also not supported

while doing group additions. Watch_group MUST always be set to OFPG_ANY for all the

group types.

32 Group table