Hi Everyone,

 

I'm only getting about 24MB/sec transfer rate between two drives, locally.  It seems this is terribly slow for two onboard disks in the same array.

 

What might be wrong?

 

Thanks,

 

Russell

Post diagnostics.

Assuming you have a parity drive enabled, that is two (or is it four?) separate array operations as its copying drive to drive.

I hope this works.  Posting by remoting with my cellphone - diagnostics attached.

 

System is parity checking and doing more than one move right now - so the diagnostics might show that.  Combined speed disk to disk, right now, is about 20MB per second.

 

Last night, just one move - no parity check - was about 30MB/sec.  It seems it should be much faster than that.

 

I don't have a parity drive - and these are Disk to Disk moves within the same array (not moving files by user shares).

 

What speed do you get for this kind of move (disk to disk, same array, Midnight Commander (local))?

 

Thanks,

 

Russell

palmwood-diagnostics-20150921-1525.zip

(I know I have some very full drives - I'm moving everything off drives 6 and 7 to retire them - and onto some new 4TB HGST NAS drives).

 

Anything strange in the diagnostics?

 

Thanks,

 

Russell

 

Finished Parity Check, drives no longer full.  Single drive to drive move - no more than 32MB/second.

 

Ideas?  (Diagnostics in other post above)

 

Thanks,

 

Russell

What speed are the drives?  Are the connect directly to says onboard or pci card? 

Also, curious, you said you don't have parity drive but just finished parity check.  Which is it?  Parity will slow down the writes.  30MB doesn't sound horrible, slow yes.

These are about the speed that I get, and after fighting it for a while, I've come to the conclusion that the problem is my Seagate NAS drives being slow.

For starters, these two statements are completely inconsistent ...

 

... System is parity checking and doing more than one move right now

 

... I don't have a parity drive - and these are Disk to Disk moves within the same array (not moving files by user shares).

 

IF you in fact have a parity protected array; and are doing more than one copy at a time, then the speed you're seeing is pretty normal.    A write operation on a parity protected array requires 4 disk I/O's ... so if the source operation is also local, then your move operation requires a LOT of disk I/O for each sector it's moving (either 8 or 9 operations).    And if you're doing more than one move at a time, the parity disk is being "thrashed" a good bit as it moves between the sectors needed for the different moves.

 

 

AND, if you're actually doing a parity check while also trying to do moves, THAT will really slow things down !!

 

I assume you didn't mean an actual parity check is in process ... but simply that there is parity involved.    But if the statement "... system is parity checking ..." was literal, that is indeed a major factor in slower speeds.

 

Thanks everyone for the help...

 

First, my mistake - of course I have a parity drive; I don't have a cache drive.  :-)  When I first posted, Parity Sync was in progress and there are probably remnants of that in my Diagnostics.  I can do a new diagnostics if that's helpful - but it sounds like my speed might be quite standard.

 

My drives are directly connected to the Motherboard - the Supermicro 2SEA/2SEE that used to be the Lime standard several years ago (low power, Celeron, 4GB RAM (I think)).  (There might be a couple on a PCI controller, but I think that's my other Unraid system.)

 

I don't know what the drive speeds are - I just assumed that 30MB/sec was slow.  It sounds slow.  I thought 5400 RPM drives were generally good for somewhere a little under 75 MB/second.

 

Maybe I have nothing to be worried about - this is standard speed?  :-)

 

Thanks,

 

Russell

30MB/sec is not slow for your setup.

Thanks Steven...

 

Just for my future shopping list...  What makes things faster?  More RAM?  More Processor?  Network?  Eliminate drives under 7200 RPM?  etc.

 

Thanks,

 

Russell

The simple fact is that writes to the parity protected array aren't going to be "fast" -- you can get up to the mid-40's on the outer cylinders with high-density (1TB/platter) 7200 rpm drives, but for most setups 30-35MB/s is about what you can reasonably expect.

 

The drive speed helps a bit -- i.e. all other things being equal 7200rpm drives will be faster than 5400/5900 rpm units -- but probably the most significant difference is using drives with high areal density.  If you move to drives with 1TB/platter density you'll see a MAJOR improvements in performance.    Your 4TB HGST drives are possibly in this category, although they makes several 5-platter 800MB/platter units as well ... the 7200rpm HDS724040ALE640  and 5700rpm HDS5C4040ALE630  both have 800MB platters.    This is still much better older drives with 500MB (or even smaller) platters.

 

If you post a picture of your Web GUI showing the drives you now have I can tell you which ones would make the most sense to move away from.

 

Odds are the RAM, processor, and network are irrelevant to your current performance => you just need to get rid of your slower drives.

 

I have an unRAID test box with a SuperMicro C2SEE, Core 2 Quad, 8GB of RAM and I get 60Mb/s+, however, I only have a few Hitachi 2TB 7200rpm drives in it.

Hmmm...  I just looked - you're getting twice the speed I am.

 

And all of my drives are 7200RPM:

 

2 x Seagate ST31500341AS / 1.5TB (retiring as soon as I get home)

4 x Seagate ST3000DM001 / 3TB

4 x Hitachi HDN724040ALE640 / 4TB

 

I guess 30MB/second is as good as it will get for now...  I'm just wondering what I could do to make it faster?  Would the RAM do it?  Something else more likely?

 

Thanks,

 

Russell

More RAM helps since unRaid caches writes (ie: its not written to disk yet, but the transfer is done), but with today's file sizes of 8Gig plus, its negligable.

 

SATA III controllers helps since just about every hard drive is SATA III, but it only helps with file sizes smaller than the cache size of the drive.

 

 

And all of my drives are 7200RPM:

 

2 x Seagate ST31500341AS / 1.5TB (retiring as soon as I get home)

4 x Seagate ST3000DM001 / 3TB

4 x Hitachi HDN724040ALE640 / 4TB

 

Of those, only the 4 Hitachi drives are likely to have decent data density. The other drives are all lower density so they're slower.

And all of my drives are 7200RPM:

 

2 x Seagate ST31500341AS / 1.5TB (retiring as soon as I get home)

4 x Seagate ST3000DM001 / 3TB

4 x Hitachi HDN724040ALE640 / 4TB

 

Of those, only the 4 Hitachi drives are likely to have decent data density. The other drives are all lower density so they're slower.

 

Actually that's not true.    The Hitachi's are only 800MB/platter.    The Seagate DM3000DM001's are 1TB/platter drives ... so they're your best drives in terms of sustained transfer speeds.    The old ST31500341AS units are by far the worst (375MB/platter) ... but you're removing those.  THAT should make a nice difference in parity check speeds !!

 

 

I have an unRAID test box with a SuperMicro C2SEE, Core 2 Quad, 8GB of RAM and I get 60Mb/s+, however, I only have a few Hitachi 2TB 7200rpm drives in it.

 

Post a screenshot of the Web GUI for your test box that shows your specific drive configuration.

 

I have an unRAID test box with a SuperMicro C2SEE, Core 2 Quad, 8GB of RAM and I get 60Mb/s+, however, I only have a few Hitachi 2TB 7200rpm drives in it.

 

Post a screenshot of the Web GUI for your test box that shows your specific drive configuration.

 

 

Sorry, but I cant right now. That server is in pieces and tucked away, and my office is in a state of total chaos right now.

 

There may be some posts around from back when rhat was my unraid server.

No problem => I was just curious what setup you might have that would explain the 60+MB/s writes you noted.

 

There are really only 4 ways to get speeds that fast ...

 

(1)  The special case of a 2-disk UnRAID [i.e. just parity and one data drive] => in this case UnRAID recognizes the configuration and it works like a RAID-1 array ... so the write penalty of 4 I/O's doesn't apply and you get VERY good write speeds.  I've seen a few cases where folks were doing this where the 2 "disks" were actually RAID-0 arrays through a controller, so there was a lot of space, but logically it was the same as if there were just 2 disks.    [bubbaQ did this a few years ago in a thread you commented in]

 

(2)  There's a cache drive, and the writes are actually going to that instead of the protected array.

 

(3)  The system has a LOT of memory (typically 32GB or more) and the writes are simply being buffered by UnRAID, which makes them SEEM a lot faster than they actually are.

 

(4)  "Turbo Write" mode is enabled, where the disk writes only 2 sequential operations instead of the normal 4.  This requires that all drives be spinning, since what actually happens is that every disk except the one being written to and the parity drive are read while the drive being written to is written  [all of these take place simultaneously, so that's 1 logical operation]; and then parity is written with the appropriate parity value [the 2nd operation].    Since this is 2 operations instead of the normal 4 operations it's appreciably faster.

 

No problem => I was just curious what setup you might have that would explain the 60+MB/s writes you noted.

 

I believe that with modern 7200rpm drives it’s easy to get 60MB/s+.

 

I get 50MB/s+ sustained writes to my array, all WD green drives and no cache, as long as I’m my writing to the first say 30 to 40% of the drive, it will slow down as the drive fills up.

 

Initial speed is 100Mb/s+ while it’s being buffered but then it’s the normal speed I get.

Yes, on a system with 1TB/platter drives you should be able to hit 50MB/s on the outer cylinders ... but NOT as a general average speed.    You'll get much higher indicated speeds at first, as the write is buffered by UnRAID, and for realtively small writes (like the graphic you posted, which appears to be about a 7-8GB write) you'll see even better speeds.  With 1TB platters and 7200 rpm you can likely even hit 60 on outer cylinder writes for perhaps the outermost 10% or so of the drives.    But again, NOT as a general average speed overall.

 

I used it as a test server, so the drives were always empty. So, I was only ever writing to the outer portion of the drives.  Maybe it was closer to 50 than 60. 

 

The point of my post was that his PCI controller and/or his drives were the limiting factor, not his motherboard.