Horton

Re: [PoC] Improve dead tuple storage for lazy vacuum

Masahiko Sawada <sawada.mshk@gmail.com>

View thread View on postgr.es
From: Masahiko Sawada <sawada.mshk@gmail.com>
To: John Naylor <john.naylor@enterprisedb.com>
Cc: Nathan Bossart <nathandbossart@gmail.com>, Andres Freund <andres@anarazel.de>, Matthias van de Meent <boekewurm+postgres@gmail.com>, Yura Sokolov <y.sokolov@postgrespro.ru>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2023-01-16T02:52:30Z
Lists: pgsql-hackers

Attachments

On Fri, Dec 23, 2022 at 4:33 PM John Naylor
<john.naylor@enterprisedb.com> wrote:
>
>
>
> On Thu, Dec 22, 2022 at 10:00 PM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>
> > If the value is a power of 2, it seems to work perfectly fine. But for
> > example if it's 700MB, the total memory exceeds the limit:
> >
> > 2*(1+2+4+8+16+32+64+128) = 510MB (72.8% of 700MB) -> keep going
> > 510 + 256 = 766MB -> stop but it exceeds the limit.
> >
> > In a more bigger case, if it's 11000MB,
> >
> > 2*(1+2+...+2048) = 8190MB (74.4%)
> > 8190 + 4096 = 12286MB
> >
> > That being said, I don't think they are not common cases. So the 75%
> > threshold seems to work fine in most cases.
>
> Thinking some more, I agree this doesn't have large practical risk, but thinking from the point of view of the community, being loose with memory limits by up to 10% is not a good precedent.

Agreed.

> Perhaps we can be clever and use 75% when the limit is a power of two and 50% otherwise. I'm skeptical of trying to be clever, and I just thought of an additional concern: We're assuming behavior of the growth in size of new DSA segments, which could possibly change. Given how allocators are typically coded, though, it seems safe to assume that they'll at most double in size.

Sounds good to me.

I've written a simple script to simulate the DSA memory usage and the
limit. The 75% limit works fine for a power of two cases, and we can
use the 60% limit for other cases (it seems we can use up to about 66%
but used 60% for safety). It would be best if we can mathematically
prove it but I could prove only the power of two cases. But the script
practically shows the 60% threshold would work for these cases.

Regards

-- 
Masahiko Sawada
Amazon Web Services: https://aws.amazon.com

Commits

  1. radixtree: Fix SIGSEGV at update of embeddable value to non-embeddable.

  2. Get rid of anonymous struct

  3. Teach radix tree to embed values at runtime

  4. Teach TID store to skip bitmap for small numbers of offsets

  5. Use bump context for TID bitmaps stored by vacuum

  6. Fix alignment of stack variable

  7. Use TidStore for dead tuple TIDs storage during lazy vacuum.

  8. Rethink create and attach APIs of shared TidStore.

  9. Fix inconsistent function prototypes with function definitions.

  10. Fix a calculation in TidStoreCreate().

  11. Fix potential integer handling issue in radixtree.h.

  12. Add TIDStore, to store sets of TIDs (ItemPointerData) efficiently.

  13. Fix link error for test_radixtree module on Windows

  14. Blind attempt to fix ODR violations

  15. Fix incorrect format specifier for int64

  16. Fix redefinition of typedefs

  17. Add template for adaptive radix tree

  18. Fix signedness error in 9f225e992 for gcc

  19. Introduce helper SIMD functions for small byte arrays

  20. Optimize vacuuming of relations with no indexes.

  21. Add bound check before bsearch() for performance

  22. Allocate consecutive blocks during parallel seqscans