Old chassis - upgrade to increase drive throughput

[cbarbie]

I have an older chassis and I am looking to increase throughput to my spinning drives.

• SuperChassis 846E1-R900B
• MB X9SCM-IIF
• Xeon E3-1275 V2
• LSI9240-8i PCI-E 2.0 x8 (in a PCI-E 3.0 x8 slot)

            speed 4GB/s

• BPN-SAS2-846EL1 – Single port

            speed  6Gb/s * 4 chan = 24Gb/s = 3GB/s (theoretical)
              3GB/s / 24 drives = 125 MB/s

 

 

Am I on the correct path with one of these two upgrades?

 

1) Go to BPN-SAS2-846EL2 – Dual port $432   (Does dual porting double the speed?)
    speed  6Gb/s * 8 chan = 48Gb/s = 6GB/s (theoretical)
    6GB/s / 24 drives = 250 MB/s
    Upgrade HBA to Pci-E 3.0 x8 (7.88GB/s)
        AOC-S2308L-L8E or equivalent

2) Go to BPN-SAS3-846EL1 – Single port $360  (will this fit in my chassis?)
    speed  12Gb/s * 4 chan = 48Gb/s = 6GB/s (theoretical)
    6GB/s / 24 drives = 250 MB/s
    Upgrade HBA to Pci-E 3.0 x8 (7.88GB/s)
        AOC-S2308L-L8E or equivalent

[JorgeB]

9 hours ago, cbarbie said:

Does dual porting double the speed?)

The difference between those models is that the EL2 has dual expanders, but that's not for speed, it's for redundancy, your current backplane already supports dual link, that doubles available bandwidth, for real world usable bandwidth examples see below:

 

https://forums.unraid.net/topic/41340-satasas-controllers-tested-real-world-max-throughput-during-parity-check/?do=findComment&comment=406521

 

[cbarbie]

Thanks for the link.  I had seen that earlier.

I currently have the HBA connected to PRI_J0.  The 2nd HBA connection goes to PRI_J1 (or PRI_J2)?  Research results were inconclusive, some said yes, some said no.

Edited August 13, 2020 by cbarbie

[JorgeB]

Usually it works with any two connections, you can also check if it's linked with single or dual link:

 

cat /sys/class/sas_host/host1/device/port-1\:0/sas_port/port-1\:0/num_phys

 

Output of 4 indicates single link, 8 dual link, command might need to be adjusted to correct host#, post diags if the above doesn't work.

[cbarbie]

Thanks for the command to check.  I had to adjust the Host and Port

Quote

root@Tower6:/# cat /sys/class/sas_host/host6/device/port-6\:0/sas_port/port-6\:0/num_phys
8

Still not seeing the speeds I think I should be getting.  A Parity check starts out at about 140MB/s (dual parity, 12 data drives, mixture of 10 & 12TB 5400&7200).   Could it be the HBA limiting it (PCI-E2.0x8)?  Is my math wrong?  Or something else?

tower6-diagnostics-20200814-0749.zip

[JorgeB]

1 hour ago, cbarbie said:

Could it be the HBA limiting it (PCI-E2.0x8)?

HBA should be capable of 2500MB/s to 3000MB/s, there's a known issue with v6.8.x where parity checks can be slower, especially for larger arrays, you could do a quick test with v6.7 to confirm if it's that.

[Vr2Io]

You have 14 disks in array

 

Aug 13 06:10:57 Tower6 kernel: unraid: allocating 72070K for 1280 stripes (14 disks)

 

By calc, 14 * 140MB/s ~ 1960MB/s, so seems dual-link still not work. Pls also confirm PCIe slot is 2.0 x8 instead x4.

Edited August 15, 2020 by Benson