Hi All,

 

I recently copied some data from one disk to another and noticed the transfer speed - 41.7 MB/sec. This was done via VM, Ubuntu Desktop 16.04 LTS.

 

It does seem slow to me but i'm not sure, and i would like to find out what speed other users are getting.

 

Also, any advice on how to improve the speed would be appreciated.

 

Thanks,

T

 

System Info:

MB: Gigabyte Z77-DS3H |

CPU: Intel® Core™ i5-3470T CPU @ 2.90GHz

HVM: Enabled

IOMMU: Enabled

Cache: 64 kB, 512 kB, 3072 kB

Memory: 16 GB (max. installable capacity 32 GB)

Network: bond0: fault-tolerance (active-backup), mtu 1500

eth0: 1000 Mb/s, full duplex, mtu 9000

eth1: not connected

tower-diagnostics-20170101-1913.zip

That's a reasonable speed for writes to the parity-protected array.  Not at all slow for the normal read/modify/write method used to write to the array.    You can get faster write speeds by using "reconstruct write" -- but this requires that all of your disks be spun up during writes.  If you're okay with that, just change the setting [settings - Disk Settings - md_write_method].

 

For going from a parity protected disk to a parity protected disk that's a reasonable (not awesome but reasonable) speed once unRaid exhausts it's memory cache for disk writes

Thanks for the info.

 

I should have stated in my previous post that I am already using the reconstruct write method. I just changed it back to read/modify/write and tried to copy a 1.6gb file, and I am getting on average 30 MB/sec.

 

Is this still a reasonable speed?

A little on the slow side if you're using modern HDDs but not by much, normal is around 40 to 50MB/s mostly dependent on the disks used.

 

Reconstruct write performs much better with writes from another computer, with disk to disk one of them will always be doing two simultaneous reads, never good for speed.

 

PS. Reconstruct write is also dependent on controller/bus bandwidth, it will never be faster than the parity check speed.

Yes, the speeds aren't bad -- as Squid noted, they're not "awesome" ... but reasonable.  If your array was all new, high-density (1TB/platter or higher) drives then I'd expect higher; but 30 or so is a reasonable expectation for most arrays.

 

I'm a bit surprised the reconstruct write didn't give better results -- but this is likely due to three factors:  (1) you're writing from a disk in the array to a disk in the array, so the reads result in extra seeks compared to writes from an external source (i.e. you'd probably get faster speeds writing from a network source); (2) the mix of your disks likely includes mixed densities; and (3) I notice you're using a pair of SASLP's ... with 8 drives attached these are limiting your bandwidth to the drives -- normally this would only be a factor with parity checks or rebuilds; but with reconstruct write you're also accessing all of the drives at once, so this would impact that as well.

 

Bottom line:  I don't think you have any issue to be concerned about.

 

 

Okay, that's good to know and again, thanks for all the info!