Not sure what changed I use a 5Gb test file from a ram disk to a test share that is set to cache only (cache is 2 240GB ssd's in raid 0) first transfer is at an steady 25MB/s then if I delete that file and subsequently transfer multiple times again it's blazing fast odd it seems backwards to me as I would think the first transfer would go to the disk cache itself any ideas what I screwed up this time 

 

 

 

Systems

My daily driver: OS Windows 10 home edition / CPU Ryzen R7 1800x 3.95GHz / Corsair H110i / PSU Thermaltake Toughpower 750watt / ASUS CH6 / Gskill Flare X 16GB DDR4 3200Mhz / HP 10GB Single Port Mellanox Connectx-2 PCI-E 10GBe NIC / Samsung 512GB 960 pro M.2 / ASUS GeForce GTX 1080 STRIX 8GB / Acer - H236HLbid 23.0" 1920x1080 60Hz Monitor x3

 

My electronics rig: OS Windows 10 home edition / CPU Ryzen R5 1600x 3.85 / Thermaltake Contac Silent 12 / PSU Thermaltake Toughpower 750watt / MSI Gaming Pro Carbon / Gskill Ripjaw 32GB DDR4 3200Mhz / HP 10GB Single Port Mellanox Connectx-2 PCI-E 10GBe NIC / X2 Sandisk 240GB SSD's / ASUS GeForce GTX 1080 STRIX 8GB / HP Monitor

 

My NAS: OS unRAID v. 6.5 rc4 / CPU Intel i7 6700 / Cooler Master MasterWatt Lite 500 Watt 80 Plus / Supermicro C7Z170-M / 16GB Samsung B die DDR4 2133Mhz / HP Mellanox ConnectX-2 10 GbE PCI-e G2 Dual SFP+ Ported Ethernet HCA NIC / 9 Drives total 17.75 TB - 1 4TB seagate parity - 4 4TB WD Red data - 1 1TB ? laptop drive data  - 2 240GB Sandisk SSD's as cache in raid 0 / Samsung PM961 256GB M.2 PCIe Internal SSD unassigned device / headless

 

 

homenas-diagnostics-20180304-1507.zip

 

These were done with 30 seconds of each other on my daily driver? Also this should not be on my 10GBe card as the path is pointing to my regular 1GBe share I named the share that when testing my network adapter

Edited March 5, 20188 yr by mrbilky
Replaced with correct zip

I'd definitely post all your questions in the Pre-Release Section. You'll get a lot more eyes there. 

https://lime-technology.com/forums/forum/65-prerelease-support/

Thought about that but it was occurring in the previous version, I posted what version I'm currently on for reference as well not sure that its actually related to upgrade was hoping for a long shot that it would be glaring in the diags I installed a SATA pcie add-in card and reconfigured and added a new drive before noticing the transfer speeds

Edited March 5, 20188 yr by mrbilky

I decided to take a look at your 'problem'.  I downloaded your diagnostics file.  First thing I noticed was that the version 6.4.1!  How do you expect anyone to help with a problem that is quite difficult to analyze when the data and the description don't match?

 

Now to get some meaningful help you are going to have to setup an experiment.   You will have to begin with a complete description of what the conditions are.  Source computer,  what the test file is (how generated), network speed, type of sdd's, configuration of sdd's, file transfer program used, transfer time.  Determine what the transfer time reported is actually is.  (Some programs report the time to transfer the data from the sending computer while others will report the total time-- i.e., including the time until the drive has written the end-of-file marker. You really would like one that reports the latter.)   

 

Then reboot the server.  Do the file transfers.  Record both the clock time and the transfer time of each transfer.  Put three minutes between each test so that anyone who looks at your log can see if there is anything significant in the log.  

 

Why is all of this information needed?  Modern OS's all use a RAM cache for  disk writes and they have ability to delay those writes for fairly long periods of time.  (That is why you have to eject USB drives before unplugging to make sure those caches have been flushed to disk!) This means that for small files  (and 5GB can be a small file depending on the amount of RAM on your server!),  the network portion of the file transfer may be done before the disk writing even starts. I would suggest using a 20GB file and change its name for each test.  (If the file header information is identical, unRAID may assume that the file is its RAM cache is the same as being transferred and use the RAM cache rather than transferring the file again.)  

 

I would actually be more concerned about your transfer speed for the first time.  I am assuming that you are actually getting 25 megabits per second (Mbs) rather than megabytes per second (MBs).  Network speed are normally measured in Mbs and drive writes speeds are measured in MBs.   (It is quite common to mix the two up and now that you realize that please make sure you type it correctly as it really does confuse people!) 

 

You should be seeing speeds in the 100Mbs range at the beginning of any file transfer to an unRAID server until the RAM cache fills.  What it drops back to at that point is the usual topic of discussion.  

 

EDIT:   You should be seeing speeds in the 100MBs range (I am assuming you r have a Gb Network)  at the beginning of your transfer.  It often drops back and that is the usual subject of discussion in most of the time.  But you said you are using ssd's in a cache drive configuration and that should have sustained write speeds close to the network speed. 

 

Edited March 5, 20188 yr by Frank1940
Brain phrart this morning

Crap I'm at work I may have accidently uploaded a previous diag file sorry bout that will post all info needed after work

Please see revised post!!!!   

Edited March 5, 20188 yr by Frank1940

With ssd's, I seem to recall, that you should be running trim...

Edited original op hoping someone will see something glaring

OK, Ihad a look and I didn't see anything in your Logs.  (Point of disclosure:  I am not a Guru on syslogs by anyone definition!) 

 

HEre is an suggestion.  I have used a program named LAN_Speed test.  you can find it here:

 

     https://totusoft.com/lanspeed

 

I used version 1.3 Lite (in the Download section on the right side of the screen).  The version does not even require installation.  Thje download is actually an .exe file and you can run it from anyplace.  (Warning---  You should check any file that you download from the Internet and verify that it is safe to use.)  However, I have been using it off and on since I downloaded it in 2016 BUT this was almost two years ago...  

 

Here is the output that I got from 5GB file that I choose to use as a test size:

 

As you can see, I had an upload speed of 90MBs to a small 'Spinner' cache drive.  If you decide to use this program, try both to your cache drive and to a Share where you are not using the cache drive.  (i.e., writing directly to the array.  You might want  to create a test share to do this but it should be easy to get rid of the share later as LANSpeed deletes the file as part of it test procedure.)  

Edited March 5, 20188 yr by Frank1940

Hum so I bought it $10 no problem thanks for the link with the crappy luck I have I will need this often but now the strange part sure looks right to me remember this is a 2 ssd cache pool in raid 0

 

 

 

 

 

 

I spent a fair period of time puzzling over your results until I suddenly realized your have a 10Gb LAN setup.  I am not prepared to very discuss what your write (and read) speed should be to your sdd's.  Someone with some experience with them will have to address that.  Remember, most folks are still using 1Gbs LAN.  And the write speeds that this test reported would support that support those data rates.

 

I notice that you were doing your transfer tests  using mapped drives  (Z:).  Let's try something different. 

 

Since I don't use Win10, I am at a bit of loss to describe how to do what I am purposing.  If you have a Network icon on the desktop, open that up and pick your server out of the choice and open that up and navigate to the point where you want to make your copy test to.   Now open up your file manager, and drop-and-drag your test file from one widow to the other.  

 

If there isn't a Network icon, Open the file manager and start looking for your Network.  It looks like this in Win7:

 

 

Do the same thing as above.  

 

 

 

 

  

5 minutes ago, Frank1940 said:

I spent a fair period of time puzzling over your results until I suddenly realized your have a 10Gb LAN setup.  I am not prepared to very discuss what your write (and read) speed should be to your sdd's.  Someone with some experience with them will have to address that.  Remember, most folks are still using 1Gbs LAN.  And the write speeds that this test reported would support that support those data rates.

 

I notice that you were doing your transfer tests  using mapped drives  (Z:).  Let's try something different. 

 

Since I don't use Win10, I am at a bit of loss to describe how to do what I am purposing.  If you have a Network icon on the desktop, open that up and pick your server out of the choice and open that up and navigate to the point where you want to make your copy test to.   Now open up your file manager, and drop-and-drag your test file from one widow to the other.  

 

If there isn't a Network icon, Open the file manager and start looking for your Network.  It looks like this in Win7:

 

 

Do the same thing as above.  

 

 

 

 

  

 

I'm game I'll work on it after work tomorrow that test share is misleading in the fact that it is actually just a share named 10Gb that is cache only on my 1Gb network at least it should not be set up on my 10Gb side of the coin, to obtain 10Gb I have to map the share via i.p. address correct? that is how I read it was done and how I did it while testing the Mellanox cards I picked up. I'm starting to wonder about the test file I'm using and if it may somehow be affecting these results I pulled it off the web as I don't have a large file to use thing is I've used it for some time now so unless it was corrupted or something I don't see how it could.