Jump to content

copy to user shares MUCH MUCH slower then copy to single disk


Recommended Posts

Posted

i have 3 3tb drives 1 with parity. Ive set up a user share, split level left blank as i just want to span across all disks. hight water allocation.

 

if i copy a movie file to the individual disks i get 100MB a second transfer rate. If i copy to the user share i get 50MB. Why is there such a huge difference? I assume when copying to the user share it will just copy it to a single disk anyways which should result in about the same speed? ive checked the disks individually and the files are kept on the same drive. I have even disabled parity for testing.

 

any suggestions?

Posted

anyone? I just dont understand why user shares would be 50% slower the disk shares! this is just to big of a difference even though its still writing to the same drive anyways!

Posted

alright so i just made a new unraid system with 3x2tb wd green hard drives on a Q6600, gigabyte ga31-m and set it up exactly the same and I get full speeds in the user shares

 

however the other system still has problems. it consists of a Asus e35m1-m cpu intergrated motherboard and 3x3TB WD green hard drives. they both are set up identical. However this one gets terrible speeds in the user share folder. Im starting to think if there is a compatibility problem or the cpu not fast enough or something?

Posted

Yup. This has always been the case.

 

I've actually dispensed with user shares for everything apart from my Media share, because performance is significantly less rather than going straight to the disks themselves.

 

I think it's on Tom's to do list, as performance, especially regarding time-macine is quite frankly... lacklustre.

 

But security over performance.

 

I should note that it is very CPU dependant. I can more than double share transfer speeds going from 800MHz to 3.2Ghz!

Posted

I should note that it is very CPU dependant. I can more than double share transfer speeds going from 800MHz to 3.2Ghz!

I would guess that is the buss performnce speed. that would not be a fair contest comparing a P3 to a sandybridge unless you had both CPU's on the same motherboard and no hardware changed other then the CPU.

 

My Atom system is just as fast, if not faster, then my Sandybridge Xeon for network speeds and writing to the array. The only time the atom cant keep up is with CPU intensive applications like PAR and RAR of massive sets.

 

 

@alfredbra:

Assuming you are using green drives, something sounds fishy that you can write to a parity protected single drive at 100MB/s while the  array caps at 50MB/s

One thing to keep in mind, Green 3TB drives do get very slow on the outer tracks (about 1/2 - 2/3 the performance of the first tracks)...

Syslog please.

Posted

@alfredbra:

Assuming you are using green drives, something sounds fishy that you can write to a parity protected single drive at 100MB/s while the  array caps at 50MB/s

 

Something seems fishy here from my perspective too.

100MB/s on a parity protected drive? I don't believe it.  

 

An empty drive's write speed on a local file system unprotected tops out at around 120MB/s-130MB/s.

 

I know with all my tweaking and RAID0 on parity on it's own caching controller with all sorts of kernel tweaks, the fastest I could get was 60MB/s at the start of the transfer.

Near the end of 10GB it gets to around 35MB/s.

 

50MB/s to an array drive on user share is a good speed. (a really good speed).

 

All depends on your benchmark.

 

Typically writes to the user share are slower, but I don't remember it being 1/2 as fast as a regular array drive.

I thought there was a 25%-33% loss in speed, unless you are using a cache drive.

 

Please check out my writeread10gb tool here.

http://code.google.com/p/unraid-weebotech/downloads/list

 

This will give you a realistic best possible case benchmark on your local machine without networking.

Posted

I should note that it is very CPU dependant. I can more than double share transfer speeds going from 800MHz to 3.2Ghz!

I would guess that is the buss performnce speed. that would not be a fair contest comparing a P3 to a sandybridge unless you had both CPU's on the same motherboard and no hardware changed other then the CPU.

 

What bus?

 

Nothing is changing apart from the CPU multiplier on my Phenom X4... [Note: Increase isn't so much when it's direct to parity, mainly cached shares]

 

But anyway, it is true that he shouldn't be getting 100MB/sec to a parity protected share without a cache drive.

 

Even 50MB/sec is too high. Most i've ever managed really is about 40MB/sec.

 

Going to change my parity drive to a WD Black and onto a separate controller next time I upgrade the machine.

Posted

transfer speeds are correct. And i have to say yes i was suprised by these speeds myself. I work at a computer shop so i have access to alot of different parts so i might try a few more different motherboard combinations. The machine is at the clients house now so I will have to try and get some syslogs from it soon. However at this point it looks like perhaps it has something to do with the CPU.

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...