[
https://issues.apache.org/jira/browse/HUDI-4368?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexey Kudinkin updated HUDI-4368:
----------------------------------
Description:
As a follow-up to the recent discussions in the community regarding
out-of-the-box configuration (DB blog), i think we should adjust some aspects
of our OOB configuration to stay in-line with other formats as it's inevitable
that people would be comparing Hudi's performance against Delta and Iceberg:
For example, we should make sure that whenever someone is creating a table from
scratch we always use "bulk_insert" instead of "upsert" as there's no reason
for us to incur the overhead of upserting since we know the table was empty.
It could roughly go as following:
* If the table is empty, and
* There's no explicit operation configured, and
* There's no pre-combining configured
Then we treat it as "bulk_insert" case
was:
As a follow-up to the recent discussions in the community regarding
out-of-the-box configuration (DB blog), i think we should adjust some aspects
of our OOB configuration to stay in-line with other formats as it's inevitable
that people would be comparing Hudi's performance against Delta and Iceberg:
For example, we should make sure that whenever someone is creating a table from
scratch we always use "bulk_insert" instead of "upsert" as there's no reason
for us to incur the overhead of upserting since we know the table was empty.
> Make sure Hudi always does bulk-insert during the first commit into the table
> -----------------------------------------------------------------------------
>
> Key: HUDI-4368
> URL: https://issues.apache.org/jira/browse/HUDI-4368
> Project: Apache Hudi
> Issue Type: Bug
> Reporter: Alexey Kudinkin
> Assignee: Raymond Xu
> Priority: Blocker
> Fix For: 0.12.0
>
>
> As a follow-up to the recent discussions in the community regarding
> out-of-the-box configuration (DB blog), i think we should adjust some aspects
> of our OOB configuration to stay in-line with other formats as it's
> inevitable that people would be comparing Hudi's performance against Delta
> and Iceberg:
> For example, we should make sure that whenever someone is creating a table
> from scratch we always use "bulk_insert" instead of "upsert" as there's no
> reason for us to incur the overhead of upserting since we know the table was
> empty.
>
> It could roughly go as following:
> * If the table is empty, and
> * There's no explicit operation configured, and
> * There's no pre-combining configured
> Then we treat it as "bulk_insert" case
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
