[
https://issues.apache.org/jira/browse/IGNITE-19505?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kirill Gusakov updated IGNITE-19505:
------------------------------------
Attachment: diag3.puml
diag2.puml
diag1.puml
flow.puml
> Design distribution zone based partition replication
> ----------------------------------------------------
>
> Key: IGNITE-19505
> URL: https://issues.apache.org/jira/browse/IGNITE-19505
> Project: Ignite
> Issue Type: Task
> Reporter: Kirill Gusakov
> Assignee: Kirill Gusakov
> Priority: Major
> Labels: ignite-3
> Attachments: diag1.puml, diag2.puml, diag3.puml, zone-collocation.md
>
>
> *Motivation*
> The current data layout is builded around the table partitions. Each table
> partition forms the group of replicas with the RAFT protocol (pluggable
> protocol in future). This layout was great up to the moment, when we decided
> to introduce the support of collocation for any chosen group of tables.
> After some rounds of discussions we decided to use the distribution zone as
> the scope of collocation. So, the partition will be managed by the
> distribution zone associated entity, instead of the table one.
> *Definition of done*
> Under this issue we need to provide the design document, which will describe
> the process for migration the current state to the:
> - Table partitions must be the part (one-to-one relation) of the zone
> partition.
> - The data layout for partitions of different tables inside the one zone
> partition - must be described as well.
> - Tables, which share the one distribution zone will be collocated (the same
> key must be stored in the same distribution zone partition)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
