-
Notifications
You must be signed in to change notification settings - Fork 409
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.
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
Change mark_cache size calculation to cut down memory usage #6861
Conversation
[REVIEW NOTIFICATION] This pull request has been approved by:
To complete the pull request process, please ask the reviewers in the list to review by filling The full list of commands accepted by this bot can be found here. Reviewer can indicate their review by submitting an approval review. |
/run-all-tests |
/run-unit-test |
/run-all-tests |
/rebuild |
/run-all-tests |
/run-sanitizer-test tsan |
@@ -116,7 +116,7 @@ class PODArrayBase : private boost::noncopyable | |||
|
|||
void alloc_for_num_elements(size_t num_elements) | |||
{ | |||
alloc(roundUpToPowerOfTwoOrZero(minimum_memory_for_elements(num_elements))); | |||
alloc(minimum_memory_for_elements(num_elements)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Plz add some comments to this method. About why don't we use roundUpToPowerOfTwoOrZero.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM with minor comments
…/tiflash into hongyunyan_mark_cache
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The reduce of memory seems to be not as notable as expected. But anyway it's an improvement. Congrats!
/merge |
@hongyunyan: It seems you want to merge this PR, I will help you trigger all the tests: /run-all-tests You only need to trigger If you have any questions about the PR merge process, please refer to pr process. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the ti-community-infra/tichi repository. |
This pull request has been accepted and is ready to merge. Commit hash: 6a35dc2
|
What problem does this PR solve?
Issue Number: close #7144
Problem Summary:
What is changed and how it works?
Currently, we calculate the mark_cache memory usage with marks.size() * sizeof(MarkInCompressedFile);
However, PODArray will do roundUpToPowerOfTwoOrZero when allocate element memory usage.
Thus, the memory usage is quite bigger then the calculation ways, which is close in twice time of the calculation ways.
To ensure the mark_cache only use the memory as we allocate, we change the memory usage calculation approach.
Then, we take the memory of allocation for MarkInCompressedFile for consideration. I think when we just initialization MarkInCompressedFile with the element number, we don't need to make the memory roundUpToPowerOfTwoOrZero. Thus, we change the alloc in PODArray(size n), to cut down the usage.
Check List
Tests
I compare the cluster with master tiflash and tiflash of this PR, to run a typical Workload.
We can find for the master tiflash, the memory will reach 13.0GB steeply and then continue slowing growth.
While for the tilfash of this PR, the memory will reach 11.5GB steeply and then continue slowing growth.
Thus, in a typical workload, we can cut down nearly 1.5GB.
Side effects
Documentation
Release note