Performance issues

Status
Not open for further replies.

pierrec

Cadet
Joined
Mar 11, 2018
Messages
2
Hello FreeNasers,

Short story : I upgrade my NAS from a RAIDZ1 with 3 disks to a RAIDZ2 with 5 disks (adding 2 disks and rebuilding my pool). The bandwidth was downgraded from 110 Mb/s (Ethernet limit) to 70 Mb/s.

Long story :
I've build a previous FreeNas with low cost hardware (16 Gb of RAM, G4400 processor, 5x 4Tb Red and IronWolf in RAIDZ2, no SSD cache) and I had a 100-110 Mb/s bandwith on rsync transfers in Samba.
I've then build a second one with better hardware (32 Gb of RAM, i5-7400, 3x 8Tb Ironwolf, no SSD cache) and had similar performance limited by Ethernet.
Then I've decided to upgrade this NAS adding 2 more disks and a SSD cache. I had to rebuild my pool and then I had to perform a lot of rsync (backup and restore). With this new setup the transfer rate is significantly slower. I've tried to remove SSD caching for no significant effect.
Also I use encryption on these 3 configurations.
I've tried to figure out the reporting provided by freeNas : the cpu remains above 40%, the disks are limited to 40 MB/s (they can perform much more then that), the interface traffic peaks at 1000M but not continuously during a transfer.

I know that RAIDZ2 is slower than RAIDZ1, but it shouldn't be the case with more disks since the bandwith is share on more disks. Any clue or advice on test to perform will be much apreciated.

Pierre
 

joeschmuck

Old Man
Moderator
Joined
May 28, 2011
Messages
10,994
The problem when testing bandwidth is you need to conduct the same exact tests and if you are just doing an RSync then it's not the same exact test, the data changes, meaning the file sizes. Here are a few questions:

1) What version of FreeNAS are you running?
2) Did you upgrade your version of FreeNAS?
3) Have you tried some real throughput tests besides RScyn?

My suggestion is to conduct an internal throughput test and then external throughput tests and see what the results are.

For internal throughput do something like dd on an uncompressed dataset. (Note: Always use uncompressed datasets for benchmark testing of your drives/storage)

For external testing you can use a SMB share and Crystal Disk Mark or any other testing you like but be consistent in your test setup so you can compare Apples to Apples.

EDIT: Many users have issues with RSync and it's speed, it may have nothing to do with your hardware setup.
 

pierrec

Cadet
Joined
Mar 11, 2018
Messages
2
1) I have the latest version. Today it's FreeNAS-11.1-U1.
2) I didn't upgrade FreeNas around this hardware upgrade. I did it before, and until.
3) No I didn't tried some real throughput tests. I will then follow your advice when I'll find a calm period on the NAS.
Thank you joe.
 

joeschmuck

Old Man
Moderator
Joined
May 28, 2011
Messages
10,994
No problem. I hope your testing shows that you do have maximum throughput and it's just RSync or small files giving you the values you are seeing.
 
Status
Not open for further replies.
Top