5 ms·

I converted from MySQL (before whole MariaDB and fork), and I've been happier with every new version. My biggest moment of joy was JSONB and it keeps getting be

I converted from MySQL (before whole MariaDB and fork), and I've been happier with every new version. My biggest moment of joy was JSONB and it keeps getting better. Can we please make the connections lighter so that I don't have to use stuff like pgbouncer in the middle? I would love to see that in future versions.

paulryanrogers 5y ago
FWIW Mysql 8 has gotten a lot better in standards compliance and ironing out legacy quirks, with some config tweaks. While my heart still belongs to PostgreSQL things like no query hints, dead tuple bloat (maybe zheap will help?), less robust replication (though getting better!), and costly connections dampens my enthusiasm.
- assface 5y ago
  > maybe zheap will help? According to Robert Haas, the zheap project is dead.
  - eurg 5y ago
    Where did Robert Haas say this? Quick search didn't surface much, except for a blog post from July by Hans-Jürgen Schönig: https://www.cybertec-postgresql.com/en/postgresql-zheap-current-status/ https://www.cybertec-postgresql.com/en/postgresql-zheap-curr... Doesn't sound like they discontinued the project.
  - Tostino 5y ago
    I'd love to see that info, been interested in why it was taken up by Cybertec and EDB seemingly dropped it.
xwdv 5y ago
Lighter connections would finally allow for using lambda functions that access a Postgres database without needing a dedicated pgbouncer server in the middle.
- jeltz 5y ago
  Yes, but this patch does not make startup cheaper, it only decreases the performance cost of concurrent open connections.
- plasma 5y ago
  Not sure if you’re using this, but AWS has RDS Proxy as a service, in case you’re hosting your own
jkatz05 5y ago
I'd be curious to see if the concurrency improvements in PostgreSQL 14 help with increasing the threshold for when you need to introduce a connection pooler.
- darksaints 5y ago
  Check out the Azure team's benchmarks. Pretty damn impressive. https://techcommunity.microsoft.com/t5/azure-database-for-postgresql/how-we-shipped-postgresql-14-on-azure-within-one-day-of-its/ba-p/2801300 https://techcommunity.microsoft.com/t5/azure-database-for-po...
darksaints 5y ago
There has been a lot of improvement in this release. https://www.depesz.com/2020/08/25/waiting-for-postgresql-14-improvements-for-handling-large-number-of-connections/ https://www.depesz.com/2020/08/25/waiting-for-postgresql-14-...
- the_duke 5y ago
  The benchmark linked in the comments shows 7-38% improvements. Nothing to scoff at, but if you need PGBouncer that probably won't make enough of a difference.
  - darksaints 5y ago
    It certainly isn't much of an improvement in connection latency (the connections are still pretty heavy), but it is a massive improvement in transaction throughput with higher numbers of connections. If you scroll down a bit, there is now higher TPS even at 5000 connections than previously could be had at 100 connections. That fixes a massive amount of pain that previously only could be solved with pgbouncer.

paulryanrogers 5y ago

FWIW Mysql 8 has gotten a lot better in standards compliance and ironing out legacy quirks, with some config tweaks. While my heart still belongs to PostgreSQL things like no query hints, dead tuple bloat (maybe zheap will help?), less robust replication (though getting better!), and costly connections dampens my enthusiasm.

assface 5y ago
> maybe zheap will help? According to Robert Haas, the zheap project is dead.
- eurg 5y ago
  Where did Robert Haas say this? Quick search didn't surface much, except for a blog post from July by Hans-Jürgen Schönig: https://www.cybertec-postgresql.com/en/postgresql-zheap-current-status/ https://www.cybertec-postgresql.com/en/postgresql-zheap-curr... Doesn't sound like they discontinued the project.
- Tostino 5y ago
  I'd love to see that info, been interested in why it was taken up by Cybertec and EDB seemingly dropped it.

xwdv 5y ago

Lighter connections would finally allow for using lambda functions that access a Postgres database without needing a dedicated pgbouncer server in the middle.

jeltz 5y ago
Yes, but this patch does not make startup cheaper, it only decreases the performance cost of concurrent open connections.
plasma 5y ago
Not sure if you’re using this, but AWS has RDS Proxy as a service, in case you’re hosting your own

jkatz05 5y ago

I'd be curious to see if the concurrency improvements in PostgreSQL 14 help with increasing the threshold for when you need to introduce a connection pooler.

darksaints 5y ago
Check out the Azure team's benchmarks. Pretty damn impressive. https://techcommunity.microsoft.com/t5/azure-database-for-postgresql/how-we-shipped-postgresql-14-on-azure-within-one-day-of-its/ba-p/2801300 https://techcommunity.microsoft.com/t5/azure-database-for-po...

darksaints 5y ago

There has been a lot of improvement in this release. https://www.depesz.com/2020/08/25/waiting-for-postgresql-14-improvements-for-handling-large-number-of-connections/ https://www.depesz.com/2020/08/25/waiting-for-postgresql-14-...

the_duke 5y ago
The benchmark linked in the comments shows 7-38% improvements. Nothing to scoff at, but if you need PGBouncer that probably won't make enough of a difference.
- darksaints 5y ago
  It certainly isn't much of an improvement in connection latency (the connections are still pretty heavy), but it is a massive improvement in transaction throughput with higher numbers of connections. If you scroll down a bit, there is now higher TPS even at 5000 connections than previously could be had at 100 connections. That fixes a massive amount of pain that previously only could be solved with pgbouncer.