Thread: regarding PG on ZFS performance

Just re-ran some older tests with the current version of Ubunutu & ZFS (but older kernel thanks to a multi-way
incompatibilitywith other things). Results are that with proper tuning, ZFS RAIDZ1 on 4 NVMe drives gives higher TPS on
pgbenchat scale 10,000 than XFS on one of the same NVMe--but the initial population of the db takes 25% longer. 

Proper tuning: PG full_page_writes off (for ZFS, on for NVMe); ZFS lz4 compression, 64K recordsize, relatime

db created by: pgbench -i -s 10000 --foreign-keys test

benchmarked as: pgbench -c 100 -j 4 -t 1000 test

NVMe: 31,804 TPS
RAIDZ1: 50,228 TPS

Some other notes:

- the situation is reversed, single NVMe is faster when using 10 connections instead of 100
- these tests are all from within containers running on Kubernetes--pg server and client in same container, connected
overdomain sockets 
- 256GB and 48 CPU pod limits--running where there's still the cgroup double-counting bug, so CPU is theoretically
throttledto ~24, leaving ~20 to PG server 
- the container is actually getting very slightly throttled at barely over 20 CPU--so not sure if it's CPU-bound or
IO-bound
- PG settings are set up for a larger database, shared_buffers, work_mem, parallel workers, autovacuum, etc
- I'd read that because of the way ZFS handles RAIDZ1 compared to RAID5, that performance probably didn't suffer
relativeto RAID10, and this is the case--tests with ZFS RAID10 on the same drives were a tiny bit slower (2-3%) than
RAIDZ1for TPS, but a bit faster on initial population (6-8%) 
- as an aside, WekaFS (https://www.aspsys.com/solutions/storage-solutions/weka-io/) is about 10% faster than RAIDZ1
(bothTPS and initial fill) 

I hope that experience from someone who actually bothered to read up on how to configure ZFS for PG can put to rest
some"ZFS is too slow" misinformation. I am certain that ZFS is not nearly the fastest for all configurations (for
instance,I am unable to configure the 4 NVMe drives into a hardware RAID10 to test, and it seems that ZFS may not scale
wellto larger numbers of disks) but "too slow to ever be consider for serious work" is flat-out wrong. 

--
Scott Ribe
scott_ribe@elevated-dev.com
https://www.linkedin.com/in/scottribe/

> On Apr 12, 2022, at 10:54 AM, Gogala, Mladen (Short Hills) <mladen.gogala@Fiserv.com> wrote:
>
> I wouldn’t call turning off data safety “a proper tuning”. It may be faster but can never be deployed to production.

As I explained, that which it protects against cannot happen on a ZFS volume. Therefore turning it off is just fine.

(ZFS fsync operations are themselves atomic, using its own write-ahead log. So full_page_writes does a double write to
avoidcorruption, but so does ZFS, thus 4X writes, and a major source of slowness. There is NO danger in letting only
oneof the two do a double-write in order to avoid a partial write ever being read back.) 

> On Apr 12, 2022, at 1:48 PM, Scott Ribe <scott_ribe@elevated-dev.com> wrote:
>
> As I explained...

Well, dang. I previously explained that in a different venue.