DDS QoS Policy Guide

A plain-language reference to the 16 DDS QoS policies that shape every ROS 2 topic.

QoS policies are the settings that decide how a topic behaves, for example whether delivery is guaranteed, whether a node that joins late still receives past messages, and how long each message stays in the queue. Every policy has its own page here that explains what it does, the values it accepts, and how it acts at each stage of a connection. Start anywhere or read straight through, then follow the links to the Dependency Map to see how policies affect one another, or to Rules & Evidence to see which combinations QoS Guard flags.

16 QoS Policies

ENTITY FACTORYENTFAC — discovery timing

PARTITIONPART — logical segmentation

USER DATAUSRDATA — participant metadata

GROUP DATAGRPDATA — pub/sub metadata

TOPIC DATATOPDATA — topic metadata

RELIABILITYRELIAB — delivery guarantee

DURABILITYDURABL — late joiners

DEADLINEDEADLN — temporal constraints

LIVELINESSLIVENS — publisher activity

HISTORYHIST — sample retention

RESOURCE LIMITSRESLIM — cache bounds

LIFESPANLFSPAN — sample validity

OWNERSHIPOWNST — multiple writers

DESTINATION ORDERDESTORD — sample ordering

WRITER DATA LIFECYCLEWDLIFE — instance cleanup

READER DATA LIFECYCLERDLIFE — purge timing

Lifecycle of DDS Communication

Three Phases: Discovery → Data Exchange → Disassociation.

1. Discovery Phase

Entities with the same topic are matched through PDP/EDP protocols and
established after verifying QoS compatibility.

2. Data Exchange Phase

Matched pairs exchange user data and control metatraffic (HEARTBEAT/ACKNACK) to
ensure reliability and timeliness.

3. Disassociation Phase

Communication ends by disposing instances or removing GUIDs, followed by
purging all history after a timeout.

Lifecycle Mapping

QoS Policy Discovery Data Exchange Disassociation

ENTITY_FACTORY O

PARTITION O

USER_DATA O

GROUP_DATA O

TOPIC_DATA O

RELIABILITY O O O

DURABILITY O O O

DEADLINE O O

LIVELINESS O O O

HISTORY O

RESOURCE_LIMITS O

LIFESPAN O

OWNERSHIP O O O

DESTINATION_ORDER O O

WRITER_DATA_LIFECYCLE O

READER_DATA_LIFECYCLE O

Summary: Metadata, Matching, Cache, and Lifecycle

Category QoS One-line summary

Discovery timing ENTFAC When entities participate in discovery

Logical segmentation PART Partition names to separate or group topic flows

Metadata USRDATA, GRPDATA, TOPDATA Application info on Participant / Pub-Sub / Topic

Delivery guarantee RELIAB best_effort vs reliable (retransmission, ordering)

Late joiners DURABL How much past data late joiners can receive

Temporal constraints DEADLN, LIVENS Period (deadline) monitoring; Publisher liveness

Cache HIST, RESLIM, LFSPAN How many samples, upper bounds, validity duration

Multiple writers OWNST, DESTORD Single owner vs shared; order by reception vs source timestamp

Instance cleanup WDLIFE, RDLIFE Dispose on unregister; when to purge disposed/no-writer samples

Next steps

Section	Use it for
Dependency Map	See how the policies affect one another across all 40 rules
Rules & Evidence	Open each rule with its conflict condition, engine checks, and the evidence behind it