nekocave.xyz

# #. Pure. #
Model Family: Seagate Constellation ES.2 (SATA 6Gb/s)

pgbench --host 127.0.0.1 --port 20001 --username pgbench --protocol=prepared -P 30 --time=300 --client=50 --jobs=200 pgbench

transaction type: <builtin: TPC-B (sort of)>
scaling factor: 1
query mode: prepared
number of clients: 50
number of threads: 50
duration: 300 s
number of transactions actually processed: 34293
latency average = 437.711 ms
latency stddev = 523.100 ms
tps = 114.138380 (including connections establishing)
tps = 114.144633 (excluding connections establishing)


PostgreSQL Docker. Tuned. Same setup
pgbench --host 127.0.0.1 --port 20002 --username pgbench --protocol=prepared -P 30 --time=300 --client=50 --jobs=200 pgbench

transaction type: <builtin: TPC-B (sort of)>
scaling factor: 1
query mode: prepared
number of clients: 50
number of threads: 50
duration: 300 s
number of transactions actually processed: 34270
latency average = 437.918 ms
latency stddev = 513.038 ms
tps = 114.077176 (including connections establishing)
tps = 114.097972 (excluding connections establishing)


#. No docker. Single # SSD 970 EVO Plus 1TB
pgbench --host 127.0.0.1 --port 5432 --username pgbench --protocol=prepared -P 30 --time=300 --client=50 --jobs=200 pgbench

transaction type: <builtin: TPC-B (sort of)>
scaling factor: 1
query mode: prepared
number of clients: 50
number of threads: 50
duration: 300 s
number of transactions actually processed: 1161892
latency average = 12.899 ms
latency stddev = 16.551 ms
tps = 3872.717477 (including connections establishing)
tps = 3876.130444 (excluding connections establishing)


:(

via https://nekocave.xyz/display/a8d7043b-9961-2392-9364-036918484829

#
Windows 10. No docker. Tuned. # WDC WD10EZEX-00WN4A0

pgbench --host 127.0.0.1 --port 5432 --username pgbench --protocol=prepared -P 30 --time=300 --client=10 --jobs=200 pgbench

transaction type: <builtin: TPC-B (sort of)>
scaling factor: 1
query mode: prepared
number of clients: 20
number of threads: 20
duration: 300 s
number of transactions actually processed: 1069747
latency average = 5.606 ms
latency stddev = 7.105 ms
tps = 3565.639504 (including connections establishing)
tps = 3567.505530 (excluding connections establishing)
@Taciturn Звонок другу: почему обычный 2TB работает в 34 раза быстрее чем Raid5 на Seagate Constellation ES.2? Это деградация от тормозных Сигейтов (с заявленными 6 GB/s) или от Raid5?

Мне пришлось дропнуть pg_restore, который длился без конца и края уже 7 дней и включить старый комп. Это невыносимо долго
Taciturn mastodon (AP)
Если и RAID5, то либо перестраивающийся, либо без одного диска. Скорее всего проблема в дисках. Я бы одному попробовал с них читать, например через DD, возможно на каком-то будет сильная просадка скорости. Ещё можно на SMART погадать, но это не точно.
первое что я сделал это cat mdstat
[email protected] ~ # cat /proc/mdstat
Personalities : [raid1] [raid6] [raid5] [raid4] [linear] [multipath] [raid0] [raid10]
md1 : active raid1 sdd2[1] sdb2[0] sda2[2] sdc2[3]
523264 blocks super 1.2 [4/4] [UUUU]

md2 : active raid5 sdb3[0] sdd3[1] sda3[2] sdc3[4]
8587496448 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/4] [UUUU]
bitmap: 6/22 pages [24KB], 65536KB chunk

md0 : active raid1 sdd1[1] sda1[2] sdc1[3] sdb1[0]
67042304 blocks super 1.2 [4/4] [UUUU]

unused devices: <none>