• • Isn't it always about speed...oh, and money.  But this time, speed.
     
    Version:  5.0-rc12a
     
    New build:
    System: ASRock - B75 Pro3
    CPU: Intel® Core i3-2100 CPU @ 3.10GHz - 3.1 GHz
    Cache: 512 kB, 64 kB, 3072 kB
    Memory: 4096 Module (max. 32 GB)
    Network: eth0: 1000Mb/s - Full Duplex
     
    Array (mostly on-board 6GB/s SATA plugs)
    Parity:  3TB Red
    Cache: 1TB Black
    Data:  1x2TB Red, 1x2TB Black, 4x2TB Green
     
    Trying to squeeze every bit of the speed out of the setup to get fast writes to the array.  I've attempted to look and try certain actions based on every Tom, Dick, and Harry's recommendation, but not overly confident on the "why" to do something other than just the simple act of trying it.
     
    test PC: Win8 HP AIO on a 100 Mbps hardline
     
    Moving 1943818KB file and reporting what Win8 says the speed obtained:
     
    
  1. 
     
  2. Array -> C:\ (WD 1TB Green EARS):  11.4MB/s
     
  3. C:\ -> Cache:  11.4MB/s
     
  4. Cache(AppData share) manually Win8 -> Array:  6.0MB/s
     
  5. C:\ -> Array:  11.4MB/s
     
  6. Cache(AppData share) manually using MC -> Disk 6 (2TB EZRX):  25.5MB/s

   

  test PC: Win7 Dell on a 1000 Mbps hardline

   

  Moving 1943818KB file and reporting what Win7 says the speed obtained:

   

  [*]Array -> C:\ (WD 80GB SE 7200):  33MB/s

  [*]C:\ -> Cache:  36MB/s

  [*]Cache(AppData share) manually Win7 -> Array:  26MB/s

  [*]C:\ -> Array:  34MB/s

   

  ---

  Do y'all think I am maxed out?  Hearing/reading about 80's and 90's...is that all because of spin speed avg for my array?

   

  Secondly, if anything is done on the array (read, write, browse) the speed of the copy dips dramatically .  I am not running Cache-dirs.

   

  Thirdly, would a syslog following a reboot help those much higher in the food chain determine if I am "tweaked"?

1. Mbps and MBps is different. Convert all to same metric.

2. Assuming nothing is using data transfer channel. Usually, the bottleneck is the hdd itself. theorically, most sata hdd(not sdd) have about 300mbps(note small case 'b'). fragmentation and physical data location contributes to the speed.

3. Physical channel quality. Having a good connection cable matters. You can measure the cable snr to get the cable performance.

 

Anyway, to cut things short, i believe there is nothing wrong with your setup. 36MBps is equals to 288Mbps.

yeah...I got the MB / mb down pat based on my previous life.

 

The #3 metric was puzzling and didn't know if the 7 vs 8 winders had issue.

 

I appreciate the net-net at the end...essentially, "quit 'cher whinin'" but I thought I was reading that array with parity may be capable of 80M (big B) speeds internally.  My Parity checks will avg. 50-70M(big B) depending...disk 2 is an older EARS that hdparm's at 50-60MB on reads.

 

 

Your Win8 PC via 100Mb/s is maxed as 100Mb/s = 12.5MB/s, and that's theoretical.  On 100Mb/s you're about as close as you'll ever get.

 

Your Win7 PC via 1000Mb/s on the other hand should be capable of much faster transfers.  1000Mb/s = 125MB/s, again theoretical.  With my setup I tend to max out around 100MB/s on reads from the array.  The drives in my system are capable of faster (they hdparm around 190MB/s), but my network tends to saturate around the 100MB/s (800Mb/s) point.  Others here report they are able to get 110 - 120 MB/s.

 

Writes to the protected array are usually going to max around 40MB/s.  Reads from the array on the other hand (or reads and writes to the non- parity protected cache) should be able to max out at the speed of the drive or network speed, whichever is lower.

 

Your issue appears to be network.  What type of cables are you using?  CAT5, 5e, 6 ,6e?  Pre-made or did you crimp them yourself?  For gigabit speeds you need CAT5e or above, CAT5 isn't capable of gigabit.  Are the server and test PC's plugged into the same gigabit switch?  Do you have a separate gigabit switch or are they plugged into a router with gigabit LAN ports?  If so, what make/model of router?  Are all devices negotiated at 1000/full?

>Do you have a separate gigabit switch or are they plugged into a router with gigabit LAN ports?

2 switches:

-Server into a NetGear Prosafe GS116 (Gigabit)

-Win7 PC into a NetGear Prosafe GS105 (Gigabit)

 

>If so, what make/model of router?

Technically 1 router and 1 Access Point:

-Router:  Linksys E3200 (gigabit ports) out to Internet

-Access Point:  Linksys WRT610n (gigabit ports)

 

 

>Are the server and test PC's plugged into the same gigabit switch?

Yes & No, the Win8 is into the GS116 & the Win7 is into the GS105 then into the GS116 "hub".  Server is in the basement into the 16port switch where all bedrooms are pulled to central access panel.  Win8 pc is wired at a desk direct to the basement.  Win7 is at the TV along with the AVR, DishNet receiver, etc...requiring more plugs hence the 5port then to the wall that is direct to the "hub" switch.

 

I could test a direct to basement connect for the Win7 temporarily bypassing the 5port, but that would ONLY be a test as I wouldn't run another pull and ultimately rely on the 5port anyway.

 

>Are all devices negotiated at 1000/full?

No on the Win8 due to internal card limitations.  Win7, assuming yes according to the ipconfig info as well as the dual lights on the switches.

 

>What type of cables are you using?Pre-made or did you crimp them yourself?

Crimped by installer during the low-volt install throughout house at time of build 8 yrs ago.  All CAT5e except for minor "jumpers" to the routers/switches which are store-bought CAT6.

 

"WD 80GB SE 7200"...  Isn't that quite old?  Possibly part of the reason that the second test was not as fast as expected?

"WD 80GB SE 7200"...  Isn't that quite old?  Possibly part of the reason that the second test was not as fast as expected?

 

Good catch S80_UK.  Without knowing the exact model number of the drive you can't look up specs.  That said, drives of that size and generation, in general, had transfer speeds around what the OP is seeing, especially mid-platter and end-platter.  It's entirely possible that the Win7 results are drive limited on the client side.

 

A good test to rule this out would be to use LAN Speed Test to copy a 1GB dummy file to/from the array and see what speeds you are getting.  Effectively rule out the drive in the Win7 machine, testing purely network.  http://www.totusoft.com/lanspeed1.html

I tried the Lite version of the link provided to test file transfer of a 1024MB file from my Win7 pc to the array.

 

Results are attached.

Those results look fine!

 

They show that read speed is good and of the same order of magnitude for both cache and array drives.  Writing to cache is faster than writing to the array although both speeds also look good.

Thanks everyone.  I'll see what I can do about swapping out the older local drive.

Thanks everyone.  I'll see what I can do about swapping out the older local drive.

Get a SSD.