[SOLVED] LSI MegaRAID SAS 9361-8i throughput

[AJAshinoff]

I have an LSI SAS 9361-8i whose sole purpose is to run a RAID 5 using 4 14TB drives as backup repository ("a place to put my stuff" Carlin) before being slowly copied offsite.

I noticed that my throughput varies between 1700 MB/min and 1100 mb/min. Oddly enough when I did a test to local drives ( three 12TB drives) attached to my Supermicro's on-board controller in the same system I'm pulling 3000+ MB/min. Both testing backup attempts used the same 2TB of data.

I wouldn't think there would be such a huge difference in throughput? Any insights?

SuperMicro Motherboard (can't see the stencil currently, there's a back up in progress and I would need to reboot to see its sign-on). I can verify that the LSI card is in a PCI3.0 x16 slot.
One Xeon E5-2620 v4 2.10GHz
128GB memory
Windows server 2019 (current)
Mirrored (onboard controller) 4Tb System Drives
Raid 5 - LSI SAS 9361-81 ( 4 WD Red 14tB Drives)
Mirrored (onboard controller ( three WD Red 12TB Drives)

 

[kanewolf]

I have an LSI SAS 9361-8i whose sole purpose is to run a RAID 5 using 4 14TB drives as backup repository ("a place to put my stuff" Carlin) before being slowly copied offsite.

I noticed that my throughput varies between 1700 MB/min and 1100 mb/min. Oddly enough when I did a test to local drives ( three 12TB drives) attached to my Supermicro's on-board controller in the same system I'm pulling 3000+ MB/min. Both testing backup attempts used the same 2TB of data.

I wouldn't think there would be such a huge difference in throughput? Any insights?

SuperMicro Motherboard (can't see the stencil currently, there's a back up in progress and I would need to reboot to see its sign-on). I can verify that the LSI card is in a PCI3.0 x16 slot.
One Xeon E5-2620 v4 2.10GHz
128GB memory
Windows server 2019 (current)
Mirrored (onboard controller) 4Tb System Drives
Raid 5 - LSI SAS 9361-81 ( 4 WD Red 14tB Drives)
Mirrored (onboard controller ( three WD Red 12TB Drives)

This review of 15TB WD Red drives -- https://nascompares.com/2019/11/15/wd-red-14tb-wd140effx-nas-hdd-early-performance-build-and-test/
shows a write throughput of 240MB/s -- without overhead, your 4 drive array could hit 1000MB/s With RAID overhead, maybe 700 - 800MB/s (guess).
So anything beyond that is memory caching in your system. A single 2TB disk can NOT get 3000MB/s. 6Gb interface limits to 500MB/s.
Your testing is flawed.

 

[AJAshinoff]

This review of 15TB WD Red drives -- https://nascompares.com/2019/11/15/wd-red-14tb-wd140effx-nas-hdd-early-performance-build-and-test/
shows a write throughput of 240MB/s -- without overhead, your 4 drive array could hit 1000MB/s With RAID overhead, maybe 700 - 800MB/s (guess).
So anything beyond that is memory caching in your system. A single 2TB disk can NOT get 3000MB/s. 6Gb interface limits to 500MB/s.
Your testing is flawed.

Thanks for the reply.

I wish that were true. Both times the numbers I saw/took were from my backup software. The 3000+ was last week with the representative from the backup software (UltraBac) remoted in and on the phone (Local SATA controller using three 12TB WD Red). The test this morning 1100 MB/min - 1700 MB/min was consistent throughout and completed successfully (RAID 5, four 14TB WD Red). The 2TB backed up (data from a server) in 1 Hour 43 min.

I am looking into the onboard memory caching. I'm also looking into the LSI RAID controller caching and whether it can be added to. With the quantity of data I need to backup it'll take a while to get the initial full backups before switching to regular incremental.

I am kinda suspicious as to why the software is reporting in MB per minute as opposed to per second.