PostgreSQL Benchmarks with pgbench and pg_test_fsync on an Intel Optane 900P 480GB PCIe SSD (where the PostgreSQL instance is running in a Docker container with the PGDATA volume and the Docker image store located on the Optane SSD)

pg_test_fsync

$ uname -a
Linux 5fc03a999611 4.15.3-2-ARCH #1 SMP PREEMPT Thu Feb 15 00:13:49 UTC 2018 x86_64 Linux
$ cat /etc/issue 
Welcome to Alpine Linux 3.6
Kernel \r on an \m (\l)
$ pg_config --version
PostgreSQL 10.1
$ cd /var/lib/postgresql/data/

$ pg_test_fsync
5 seconds per test
O_DIRECT supported on this platform for open_datasync and open_sync.

Compare file sync methods using one 8kB write:
(in wal_sync_method preference order, except fdatasync is Linux's default)
        open_datasync                     33658.059 ops/sec      30 usecs/op
        fdatasync                         32739.266 ops/sec      31 usecs/op
        fsync                             30190.285 ops/sec      33 usecs/op
        fsync_writethrough                              n/a
        open_sync                         30939.870 ops/sec      32 usecs/op

Compare file sync methods using two 8kB writes:
(in wal_sync_method preference order, except fdatasync is Linux's default)
        open_datasync                     20151.027 ops/sec      50 usecs/op
        fdatasync                         21399.872 ops/sec      47 usecs/op
        fsync                             20932.908 ops/sec      48 usecs/op
        fsync_writethrough                              n/a
        open_sync                         19837.057 ops/sec      50 usecs/op

Compare open_sync with different write sizes:
(This is designed to compare the cost of writing 16kB in different write
open_sync sizes.)
         1 * 16kB open_sync write         25223.839 ops/sec      40 usecs/op
         2 *  8kB open_sync writes        19772.462 ops/sec      51 usecs/op
         4 *  4kB open_sync writes        10238.681 ops/sec      98 usecs/op
         8 *  2kB open_sync writes         3439.561 ops/sec     291 usecs/op
        16 *  1kB open_sync writes         1603.666 ops/sec     624 usecs/op

Test if fsync on non-write file descriptor is honored:
(If the times are similar, fsync() can sync data written on a different
descriptor.)
        write, fsync, close               26943.298 ops/sec      37 usecs/op
        write, close, fsync               27315.274 ops/sec      37 usecs/op

Non-sync'ed 8kB writes:
        write                            365758.683 ops/sec       3 usecs/op

pgbench

$ uname -a
Linux 5fc03a999611 4.15.3-2-ARCH #1 SMP PREEMPT Thu Feb 15 00:13:49 UTC 2018 x86_64 Linux
$ cat /etc/issue 
Welcome to Alpine Linux 3.6
Kernel \r on an \m (\l)
$ pg_config --version
PostgreSQL 10.1

$ pgbench -i -h localhost -p 5432 -U postgres -s 25

$ pgbench -h localhost -p 5432 -U postgres -t 1000
starting vacuum...end.
transaction type: <builtin: TPC-B (sort of)>
scaling factor: 1
query mode: simple
number of clients: 1
number of threads: 1
number of transactions per client: 1000
number of transactions actually processed: 1000/1000
latency average = 0.362 ms
tps = 2759.622825 (including connections establishing)
tps = 2767.974113 (excluding connections establishing)

$ pgbench -h localhost -p 5432 -U postgres -t 1000
starting vacuum...end.
transaction type: <builtin: TPC-B (sort of)>
scaling factor: 1
query mode: simple
number of clients: 90
number of threads: 1
number of transactions per client: 20
number of transactions actually processed: 1800/1800
latency average = 31.331 ms
tps = 2872.539308 (including connections establishing)
tps = 2880.832617 (excluding connections establishing)