On Thu, Dec 27, 2018 at 02:44:55PM -0200, Haroldo Kerry wrote:
> PostgreSQL 9.6.10 on x86_64-pc-linux-gnu (Debian 9.6.10-1.pgdg80+1),
> Connected to SAN: Dell Compellent SC2020, with 7 x Samsung PM1633 SSDs
> https://www.samsung.com/us/labs/pdfs/collateral/pm1633-prodoverview-2015.pdf,
> RAID10+RAID5 configuration, 8GB Cache, read-write battery backed cache
> enabled, connected via dedicated iSCSI switches and dedicated Ethernet
> ports, in link aggregation mode (2x1Gbps max bandwidth).
> I’m trying to figure out if PostgreSQL (PG) has some inherent limit on IOPS
> per connection.
postgres uses one server backend per client.
> We tried to increase effective_io_size from 1 to 30, to no effect on
> multiple tests.
https://www.postgresql.org/docs/current/runtime-config-resource.html#RUNTIME-CONFIG-RESOURCE-ASYNC-BEHAVIOR
=> "Currently, this setting only affects bitmap heap scans."
> Is this some inherent limitation of PG or am I misunderstanding something?
It is a hsitoric limitation, but nowadays there's parallel query, which uses
2ndary "backend worker" processes.
It's supported in v9.6 but much more versatile in v10 and v11.
Justin
