Skip to content

Comments

Reduce parquet metadata read request count#15421

Merged
zhenxiao merged 1 commit intoprestodb:masterfrom
pettyjamesm:parquet-footer-reads
Nov 12, 2020
Merged

Reduce parquet metadata read request count#15421
zhenxiao merged 1 commit intoprestodb:masterfrom
pettyjamesm:parquet-footer-reads

Conversation

@pettyjamesm
Copy link
Contributor

Attempts to reduce the number of reads required for parquet footers by pre-reading up to 16KiB from the end of the file. If the metadata section ends up being < 16KiB long, then one additional round trip has been saved. When the metadata section is longer, the additional read occurs to load the remaining portion of the metadata section.

== RELEASE NOTES ==
Hive Changes
* Improve parquet metadata reader by pre-loading 16KiB from the end of the file. When the metadata section is smaller than the prefetch size, only a single read is performed compared to the previous two.

@pettyjamesm pettyjamesm requested a review from zhenxiao November 10, 2020 21:06
Copy link
Collaborator

@zhenxiao zhenxiao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

hi @pettyjamesm nice work
looks good to me. Just a few minor issues

Attempts to reduce the number of reads required for parquet footers
by pre-reading up to 16KiB from the end of the file. If the metadata
section ends up being < 16KiB long, then one additional round trip
has been saved. When the metadata section is longer, the additional
read occurs to load the remaining portion of the metadata section.
@pettyjamesm
Copy link
Contributor Author

@zhenxiao Addressed your prior feedback, ready for re-review.

@zhenxiao zhenxiao merged commit 1b84222 into prestodb:master Nov 12, 2020
@pettyjamesm pettyjamesm deleted the parquet-footer-reads branch November 12, 2020 20:58
@caithagoras caithagoras mentioned this pull request Dec 4, 2020
1 task
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants