Performance / memory usage fix: do not allocate memory in pg_tde_slot #265
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dAdAbird
approved these changes
Aug 26, 2024
|
Overall, the PR looks good. Just one minor comment. I believe we can now get rid of TdeSlotForgetDecryptedTuple function as we do not need to pfree the decrypted_tuple |
codeforall
approved these changes
Aug 27, 2024
Until now, for some reason we allocated memory for each decrypted tuple in pg_tde_slot, and only freed all that memory when the transaction ended. This caused two issues: * palloc takes time, and the many small pallocs resulted in a significant performance drop in scans * sequential scans could potentially allocate memory to the entire table, requiring way too much memory for large tables This allocation is most likely a leftover from before we even used slots, and handled selects differently. With slots, we don't need them at all, as slots are not expected to handle multiple tuples at the same time, only the last one. This commit removes the palloc completely, and instead adds a single BLCKSZ array to the slot structure, which can hold any size of decrypted tuple. To be safe, it also disables the get tuple function, forcing the core code to use copy instead when needed. This change results in: 1. no "memory spike" during sequential scans 2. ~1.55x overhead instead of ~2.2x (The TODO in slot_copytuple will be addressed in a separate commit)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Until now, for some reason we allocated memory for each decrypted tuple in pg_tde_slot, and only freed all that memory when the transaction ended. This caused two issues:
This allocation is most likely a leftover from before we even used slots, and handled selects differently. With slots, we don't need them at all, as slots are not expected to handle multiple tuples at the same time, only the last one.
This commit removes the palloc completely, and instead adds a single BLCKSZ array to the slot structure, which can hold any size of decrypted tuple.
To be safe, it also disables the get tuple function, forcing the core code to use copy instead when needed.
This change results in:
(The TODO in slot_copytuple will be addressed in a separate commit)