hey,

 

i use 2 unraid servers at the moment. the "big" one with unraid 4.7 with 15 bays (currently 10 are being used), has a read speed (copying from a workstation to the pc via lan) of 30mb/s.

 

my microserver n36l using unraid5 has a read speed of about 49mb/s. (4 slots used, 3tb drives)

 

none of them use a cache drive. did unraid5 improve the read/write speeds or is something wrong on my unraid 4 server?

 

p1lot

Maybe depends on how they are connected, which part of the HDD (inside or outside of the platter). What HDDs? 7200 vs 5000?

 

Josh

btw, the 2 servers reside in different locations.

 

big server uses 2tb drives only (hitachi 5k3000s and wd20ears)

 

small one only uses 3tb wd30ezrx.

 

 

both servers are connected to a gigabit switch. i only use cat5e lan cables.

btw, the 2 servers reside in different locations.

 

big server uses 2tb drives only (hitachi 5k3000s and wd20ears)

 

small one only uses 3tb wd30ezrx.

 

 

both servers are connected to a gigabit switch. i only use cat5e lan cables.

Small server uses faster drives.  Since the 3tb drives have a higher density of data they will likely read faster than the 2TB drives.  So it is probably just a perceived difference rather than an actual difference.  The test would be to setup the small server with the same mix of 2TB drives on unraid5 and see what your read speeds are.

That must be just  perceived, because I am running 4.7 with 9 discs all 2TB Samsungs, no cache drive and when I copy from Workstation to the array using TeraCopy it is as low as 24 MB/s and as high as 50 MB/s. Most of the time it is between 42 MB/s and 46 MB/s. Also what determines your speed is your NIC. I got a new Intel NIC and speed improved. In addition how full are your drives? When they are empty they write to the outer rings, when they are full it is being written to the inner rings which again means slower speed because of head travel time.

Only a few of the disks are half full, 2-3 are totally empty.

 

I use an ASRock mainboard with a Realtek NIC, which should be fine. Maybe the switch is the culprit? I will try another one when I have some more time.

 

Thanks for your input.

Also what determines your speed is your NIC. I got a new Intel NIC and speed improved.

 

I'm using $29 GIG-E switches and on-board Realtek / Marvell PCI-Express NIC's and hitting up to 115MB/sec read and 90MB/sec writes.  Had to use the latest drivers on the Windows machines and a mix of mostly JUMBO DISABLED, JUMBO 2k and JUMBO 3k in order to get them moving that fast.  Even through double $29 GIG-E switches I can pull around 85-95MB/sec read (haven't tested writes through two of them yet)

 

ADDED: I find my biggest problem to be my source drives.  If they are in USB, forget it only getting around 30MB/sec.  If they are on my Intel Matrix RAID-1 where I download too, I can pull 60MB/sec if its idle or 30-50MB/sec if I have downloads hitting the same array.  From my 300GB Velociraptor over GIG-E to the Samsung F4 2TB cache drive we are talking 90MB/sec writes and 115MB/sec reads.

 

ALSO, if I move jumbo up to 4-8k it really kills my numbers (and both switches say they support 8k jumbos)