Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Iceberg metadata super optimization #43460

Open
1 of 9 tasks
stephen-shelby opened this issue Apr 1, 2024 · 0 comments · Fixed by #44527, #44581, #44632, #44679 or #45640
Open
1 of 9 tasks

Iceberg metadata super optimization #43460

stephen-shelby opened this issue Apr 1, 2024 · 0 comments · Fixed by #44527, #44581, #44632, #44679 or #45640
Labels
type/enhancement Make an enhancement to StarRocks version:3.3.0

Comments

@stephen-shelby
Copy link
Contributor

stephen-shelby commented Apr 1, 2024 •
edited by wanpengfei-git

Enhancement

This issue is used to trace iceberg metadata related optimization patches. The first version patch is #43459

There are currently four important optimizations:

  1. iceberg metadata distributed plan. （Performance increased by n ~ 2n times. n is be number）
  2. iceberg manifest cache. (Job planning delay reduced to around 100ms)
  3. external table based on metadata file distributed plan Framework.
  4. refactor some iceberg metadata parts.

There is still a lot of optimization work to come, some of which are as follows:

  • Add more auto distributed strategies baed on be load like cpu/io/mem
  • Enhancement time travel statement like as of.
  • refactor current split structure like RemoteFileInfo
  • provide default optimal distributed plan threshold by do a lot of testing.
  • Adapt to partition data parser when exists partition evolution.
  • Continuously optimize performance
  • add cache metrics like hitrate
  • provide benchmark to guarantee performance
@stephen-shelby stephen-shelby added the type/enhancement Make an enhancement to StarRocks label Apr 1, 2024
@stephen-shelby stephen-shelby mentioned this issue Apr 1, 2024
Closed
24 tasks
@Dshadowzh Dshadowzh mentioned this issue Apr 12, 2024
Open
60 tasks
This was referenced Apr 17, 2024
Merged
Merged
Merged
Merged
Merged
Merged
Merged
Merged
Merged
Merged
Merged
Merged
Merged
Merged
@stephen-shelby stephen-shelby linked a pull request Apr 22, 2024 that will close this issue
[Feature] Introduce meta spec interface #44527
Merged
24 tasks
@stephen-shelby stephen-shelby mentioned this issue Apr 22, 2024
Merged
24 tasks
@andyziye andyziye reopened this Apr 23, 2024
@stephen-shelby stephen-shelby mentioned this issue Apr 23, 2024
Merged
24 tasks
@stephen-shelby stephen-shelby linked a pull request Apr 23, 2024 that will close this issue
[Enhancement] Implement iceberg metadata scan node #44581
Merged
24 tasks
@stephen-shelby stephen-shelby mentioned this issue Apr 23, 2024
Merged
24 tasks
@stephen-shelby stephen-shelby linked a pull request Apr 23, 2024 that will close this issue
[Enhancement] iceberg metadata reader execution #44632
Merged
24 tasks
This was referenced Apr 24, 2024
Merged
Merged
@stephen-shelby stephen-shelby linked a pull request Apr 24, 2024 that will close this issue
[Enhancement] support metadata collect job #44679
Merged
24 tasks
@stephen-shelby stephen-shelby mentioned this issue Apr 24, 2024
Merged
24 tasks
This was referenced Apr 26, 2024
Merged
Merged
Merged
Merged
This was referenced May 6, 2024
Merged
Merged
Merged
Merged
Merged
Merged
Merged
imay pushed a commit that referenced this issue May 11, 2024
@stephen-shelby
[Enhancement] pass connect context to the prepareMetadata and distrib…
e829f98 
…uted_plan interface (#45404)

Why I'm doing:
we use ConnectContext.get() to get Session variable in the iceberg job planning. If we don't execute planning in the query thread, and can't get the connect context. There are two cases:

In the prepareMetadata, the job planning is executed in the thread pool
distributed plan mode, the metadata collect job is executed in the another thread.
1 && 2. plan_mode is distributed and running under the thread pool
So we need to adapt to the two case with the connect context.

What I'm doing:
Adapting ConnectContext and Tracers under multi-threading

Fixes #issue
#43460

Signed-off-by: stephen <stephen5217@163.com>
stephen-shelby added a commit to stephen-shelby/starrocks that referenced this issue May 15, 2024
@stephen-shelby
[Enhancement] pass connect context to the prepareMetadata and distrib…
487de4a 
…uted_plan interface (StarRocks#45404)

Why I'm doing:
we use ConnectContext.get() to get Session variable in the iceberg job planning. If we don't execute planning in the query thread, and can't get the connect context. There are two cases:

In the prepareMetadata, the job planning is executed in the thread pool
distributed plan mode, the metadata collect job is executed in the another thread.
1 && 2. plan_mode is distributed and running under the thread pool
So we need to adapt to the two case with the connect context.

What I'm doing:
Adapting ConnectContext and Tracers under multi-threading

Fixes #issue
StarRocks#43460

Signed-off-by: stephen <stephen5217@163.com>
This was referenced May 15, 2024
Closed
Merged
@stephen-shelby stephen-shelby linked a pull request May 15, 2024 that will close this issue
[Feature] iceberg metadata super optimize (backport #45479) #45640
Merged
24 tasks
node pushed a commit to vivo/starrocks that referenced this issue May 17, 2024
@stephen-shelby @node
[Enhancement] pass connect context to the prepareMetadata and distrib…
5c3bfbe 
…uted_plan interface (StarRocks#45404)

Why I'm doing:
we use ConnectContext.get() to get Session variable in the iceberg job planning. If we don't execute planning in the query thread, and can't get the connect context. There are two cases:

In the prepareMetadata, the job planning is executed in the thread pool
distributed plan mode, the metadata collect job is executed in the another thread.
1 && 2. plan_mode is distributed and running under the thread pool
So we need to adapt to the two case with the connect context.

What I'm doing:
Adapting ConnectContext and Tracers under multi-threading

Fixes #issue
StarRocks#43460

Signed-off-by: stephen <stephen5217@163.com>
@Samrose-Ahmed Samrose-Ahmed mentioned this issue Jun 2, 2024
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
type/enhancement Make an enhancement to StarRocks version:3.3.0
Projects
None yet
Milestone
No milestone
3 participants
@wanpengfei-git @stephen-shelby @andyziye