subtransaction performance

Started by Zhihong Yuover 3 years ago4 messages
#1Zhihong Yu
zyu@yugabyte.com

Hi,
I stumbled over:

https://about.gitlab.com/blog/2021/09/29/why-we-spent-the-last-month-eliminating-postgresql-subtransactions/

I wonder if SAVEPOINT / subtransaction performance has been boosted since
the blog was written.

Cheers

#2Bruce Momjian
bruce@momjian.us
In reply to: Zhihong Yu (#1)
Re: subtransaction performance

On Fri, Oct 7, 2022 at 03:23:27PM -0700, Zhihong Yu wrote:

Hi,
I stumbled over:

https://about.gitlab.com/blog/2021/09/29/
why-we-spent-the-last-month-eliminating-postgresql-subtransactions/

I wonder if SAVEPOINT / subtransaction performance has been boosted since the
blog was written.

No, I have not seen any changes in this area since then. Seems there
are two problems --- the 64 cache per session and the 64k on the
replica. In both cases, it seems sizing is not optimal, but sizing is
never optimal. I guess we can look at allowing manual size adjustment,
automatic size adjustment, or a different approach that is more graceful
for larger savepoint workloads.

--
Bruce Momjian <bruce@momjian.us> https://momjian.us
EDB https://enterprisedb.com

Indecision is a decision. Inaction is an action. Mark Batterson

#3Nathan Bossart
nathandbossart@gmail.com
In reply to: Bruce Momjian (#2)
Re: subtransaction performance

On Mon, Oct 10, 2022 at 02:20:37PM -0400, Bruce Momjian wrote:

On Fri, Oct 7, 2022 at 03:23:27PM -0700, Zhihong Yu wrote:

I wonder if SAVEPOINT /�subtransaction performance has been boosted since the
blog was written.

No, I have not seen any changes in this area since then. Seems there
are two problems --- the 64 cache per session and the 64k on the
replica. In both cases, it seems sizing is not optimal, but sizing is
never optimal. I guess we can look at allowing manual size adjustment,
automatic size adjustment, or a different approach that is more graceful
for larger savepoint workloads.

I believe the following commitfest entries might be relevant to this
discussion:

https://commitfest.postgresql.org/39/2627/
https://commitfest.postgresql.org/39/3514/
https://commitfest.postgresql.org/39/3806/

--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com

#4Bruce Momjian
bruce@momjian.us
In reply to: Nathan Bossart (#3)
Re: subtransaction performance

On Mon, Oct 10, 2022 at 08:34:33PM -0700, Nathan Bossart wrote:

On Mon, Oct 10, 2022 at 02:20:37PM -0400, Bruce Momjian wrote:

On Fri, Oct 7, 2022 at 03:23:27PM -0700, Zhihong Yu wrote:

I wonder if SAVEPOINT / subtransaction performance has been boosted since the
blog was written.

No, I have not seen any changes in this area since then. Seems there
are two problems --- the 64 cache per session and the 64k on the
replica. In both cases, it seems sizing is not optimal, but sizing is
never optimal. I guess we can look at allowing manual size adjustment,
automatic size adjustment, or a different approach that is more graceful
for larger savepoint workloads.

I believe the following commitfest entries might be relevant to this
discussion:

https://commitfest.postgresql.org/39/2627/
https://commitfest.postgresql.org/39/3514/
https://commitfest.postgresql.org/39/3806/

Wow, odd that I missed those. Yes, they are very relevant. :-)
The only other idea I had was to report such overflows, but these are
better.

--
Bruce Momjian <bruce@momjian.us> https://momjian.us
EDB https://enterprisedb.com

Indecision is a decision. Inaction is an action. Mark Batterson