High CPU usage when transferring files locally.

[jpfordtc]

Good day, I switched to UNRAID about a year ago, and I have been mostly happy with it, however I am experiencing issues that I suspect are caused by my unusual configuration. I hope the community can help.

 

I am experiencing a strange issue that is beyond my ability to troubleshoot. As I was copying (not cutting/paste) files that are already on the NAS (in the Download folder, no cache) to their final final resting place on a cached share on the same server. The individual files being copied are 100mb to 400mb between two SMB Shares from a Windows 10 host. While copying file, the transfer speed is in the 20-30MB/sec for a single transfer, and goes way down if I start more than 1 transfer at the time. Also at the time of those transfer, I notice that the CPU usage will go into the high 90%. I was doing this at the wee hours of the morning, so no other users were using any other resources, or running other transfers. When file transfers are not running, the server idles at about 20%. The NAS runs a few Dockers (Plex, SFTP, Transmission_VPN), but all of them were idling at the time).

 

I suspect that there is something wrong going on as an internal transfer to a cached drive feels like it should be much faster even if I initiated it remotely. I also suspect this could be caused by my choice of SATA connectivity, which is split between two onboard controllers, and two expansion card. However this configuration was chosen because in my experience it is the one that had the best performance. All the details are indicated below. Also, the speed benchmark from the DiskSpeed docker indicates that my 2TB drives, 3TB, 4TB drives under-perform at approx 42MB/sec to 60MB/sec despite using different controllers. I also realize that some of the drives are not ideal for my use case, however performance seems far below what I feel like I should be experiencing.

 

My questions are:

1. Is it something with my configuration or my choice of drives?

2. Would replacing my current mix of SATA cards with a proper LSI SAS card in IT mode solve my issue?

 

 

Full configuration details below.

 

Configuration

ASUSTeK Computer INC. P6T Version Rev 1.xx

Intel® Core™ i7 CPU 920 @ 2.67GHz

12 GiB Other Multi-bit ECC

 

Drives:

2x 8TB Parity Drives

8x Drives of all sorts taken from other working equipment (1x 8TB, 2x 6TB, 1x 4TB, 1x 3TB, 3x 2TB)

Cache: 2x 500gb Sata TS480GSSD220S

 

How are the drives connected:

Here is additional data from the DiskSpeed Docker:

82801JI (ICH10 Family) SATA AHCI Controller (Onboard)

Port 1: sdc 480GB TS480GSSD220S  (Cache)

Port 2: sdd 480GB TS480GSSD220S  (Cache 2)

Port 3: sde 2TB Seagate ST2000LM015 9 (Disk 8 )

Port 4: N/A (this port is defective I think)

Port 5: sdf 4TB Western Digital WD40EZRX (Disk 3)

Port 6: sdg 8TB Seagate ST8000VN004 (Disk 5)

JMB363 SATA/IDE Controller (also onboard)

Port 1: sdm 8TB Seagate ST8000VN0022 Rev 0956 Serial: ZA1ES9NJ (Parity 2)

Port 2: N/A (I also think that this one is defective)

Device 9215 (8 ports all functional)

Port 1: sdh 2TB Seagate ST2000LM015 (Disk 6)

Port 2: sdi 6TB Western Digital WD6001FZWX  (Disk 4)

Port 3: sdj 6TB Western Digital WD60EZRX (Disk 2)

Port 4: sdk 2TB Western Digital WD20EARS (Disk 1)

Port 5: sdl 8TB Seagate ST8000VN004 Rev SC60 Serial: WKD1Q9XC (Parity)

SiI 3114 [SATALink/SATARaid] Serial ATA Controller

Port 1: N/A (only 3x ports functional)

Port 2: sdb 3TB Western Digital WD30EZRS (Disk 7)

Port 3: N/A

Port 4: N/A

 

Drive performance graph: https://bayimg.com/paOMfaAgJ

 

Thank you for your help.

 

 

 

 

 

 

Edited February 4, 2021 by jpfordtc
Edit: Smiley that shouldnt be.

[ChatNoir]

You might want to attach your diagnostics that cover a period when you experience this to your next post (Tools/Diagnostics).

 

Did you always had this issue ?

 

It might be an issue with the controller.

But it might also be that the Parity2 calculation have no hardware acceleration on your CPU and have a big impact on performance ? (no expert here )

[jpfordtc]

I have attached diagnostics as per your request.

serverv3-diagnostics-20210204-1451.zip

But yes, I have always had some performance issues. The LSI perfomances where bad when I included more than 1 drive, so I downgraded to only one. As for your Parity2 point, I always ignored that since according to the benchmark, it is performing well, but you could very well be right.

 

As for hardware acceleration, how does that work if I replace the current set of cards with a LSI 16i card. Does the hardware acceleration come from the LSI card if does it still take place on the (Northbridge or is it Southbridge)? Can the chipset handle the load or is it a non issue if I go the hardware route.

 

Edit: Typos.

 

 

Edited February 4, 2021 by jpfordtc

[ChatNoir]

If I understand that correctly, the Parity calculation is done by the CPU. And without hardware acceleration in the CPU, this can have quite an impact, particulary for Parity 2 as it uses a different algoritm.

But someone with more knowledge might correct me.