Hello there. My first post. I have recently installed an unRaid machine - 2x 2TB Hdd, 1x 3TB Parity and a 1x250GB SSD Cache. Working great, but I have noticed slow transfers speeds. Two ethernet interfaces are bonded (integrated + Intel PCI card).

 

Have been observing CPU usage, and it appears that only one of the 4 is doing all the work (> 96% utilisation), the others ~ 10 %. Using Cache Disk. Could there be an element of the configuration that is forcing single threaded behaviour - for say the cache copy - or otherwise?

 

I am using a J2900 processor - Intel embedded - in an Asrock Q2900 motherboard. Rest of system works great - have installed Plex and exploring Docker. Can you help or suggest some pointers?

Have you tried it without bonding?

UnRAID SHOULD be using all of the cores without issue.  I wonder (as trurl noted) if the network bonding is causing a single core to be bound to network.    Agree you should try this without network bonding and see if that helps.

 

Be sure you use your best NIC ... probably the Intel card as long as it's a Gb card.

 

Thank you for the fast reply!

 

Bonding disabled - using the Intel NIC. Copying from macOS still slow. Used iperf3 from hosted docker - getting 970 mbps on 1 Gbps link so all good with IO. Most curiously htop - when logging in as root over SSH, shows uniform CPU use - with no spikes. Perhaps there is something wrong with my GUI - will investigate further for root cause of slow transfer - perhaps may try jumbo frames too.

 

Screenshot of htop attached.

 

Anything further I should try?

 

Matt

 

Are you using SMB, AFP or NFS? Which version of unRAID? Which version of macOS? Some suggestions here.

Are you using SMB, AFP or NFS? Which version of unRAID? Which version of macOS? Some suggestions here.

 

And what do you mean by slow.  If you are copying a lot of small files the file creation and disk space allocation can consumed a large percentage of the total time to create, copy and store a large quantity of small files.