From: | Michail Nikolaev <michail(dot)nikolaev(at)gmail(dot)com> |
---|---|
To: | Andrey Borodin <x4mmm(at)yandex-team(dot)ru> |
Cc: | Simon Riggs <simon(dot)riggs(at)enterprisedb(dot)com>, Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com>, Alexander Korotkov <aekorotkov(at)gmail(dot)com>, Andres Freund <andres(at)anarazel(dot)de>, reshkekirill <reshkekirill(at)gmail(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Slow standby snapshot |
Date: | 2022-08-07 19:28:36 |
Message-ID: | CANtu0ojtCRHgjaFAUC8DgjrtHEwXNKPqxeoNUd1JFy8Fzmy5AQ@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Hello, everyone.
> It seems to me storing the index itself is simpler and maybe faster by
> the cycles to perform addition.
Done in v7.
> Since each xid in the tree is always stored to the right, it should be
> possible to make that significantly better by starting each binary
> search from the next element, rather than the start of the array.
Also, looks like it is better to go with `tail = Max(start,
pArray->tailKnownAssignedXids)` (in v1-0001-TODO.patch)
Performance tests show Simon's approach solves the issue without
significant difference in performance comparing to my version.
I need some additional time to provide statistically significant best
coefficients (how often to go compression, minimum number of invalid
xids to start compression).
Thanks,
Michail.
Attachment | Content-Type | Size |
---|---|---|
v7-0001-Currently-KnownAssignedXidsGetAndSetXmin-requires.patch | text/plain | 6.9 KB |
image.png | image/png | 181.8 KB |
image.png | image/png | 561.7 KB |
v1-0001-TODO.patch | text/plain | 6.7 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | Thomas Munro | 2022-08-07 21:09:53 | Re: logical decoding and replication of sequences |
Previous Message | Noah Misch | 2022-08-07 19:12:05 | Re: logical decoding and replication of sequences |