Jump to content
allischalmersman

Typical gigabit transfer speeds?

64 posts in this topic Last Reply

Recommended Posts

Ok.  As I thought, it isn't the PC.  And yes I have checked for drivers.  Here is me transferring from the desktop upstairs, to a laptop plugged in to the same switch the tower is plugged into down in the basement.  So what within Unraid or the hardware other than the PCI NIC can cause this? 

Kims PC.JPG

Share this post


Link to post

Have another suggestion for you to try.  Go to 'Settings'  >>>  'Disk Settings' and set the " Tunable (md_write_method): " to 'reconstruct write'.   Then either (1) turn off caching or (2) setup up a temporary user share (without caching) and see if the copy speeds change.  (You will be writing directly to the array with a faster write method that requires that all of the disks be spun up.  I don't have a cache drive setup on either of my servers and I get write speeds of 70+Mbps on large-size files.  With a gaggle of small-size files, it drops significantly!)  

Share this post


Link to post

Well I kind of did that. I turned on reconstruct write, and turned off cache for that share and the results were the same a few posts up the picture labeled "no cache"

No trees were harmed in the sending of this message, however, a significant number of electrons were terribly inconvenienced.

Share this post


Link to post
Install this plugin, there are few NIC related settings that can make a difference in some cases:

 

 

I already have that and flow control is turned off as is LSO. Thanks though. Really looking forward to figuring this out.

 

No trees were harmed in the sending of this message, however, a significant number of electrons were terribly inconvenienced.

 

 

 

Share this post


Link to post

Try doing the test in the other direction. Do performance speed tests copying FROM unraid to another computer.

Share this post


Link to post
Out of ideas, my guess still is something network related.

I dont disagree, but it has to be something within the Unraid box software or hardware, and it isn't the NIC. I've tried 3

No trees were harmed in the sending of this message, however, a significant number of electrons were terribly inconvenienced.

Share this post


Link to post
Try doing the test in the other direction. Do performance speed tests copying FROM unraid to another computer.

I've done that its about the same. I'll do that test after parity check completes.

No trees were harmed in the sending of this message, however, a significant number of electrons were terribly inconvenienced.

Share this post


Link to post

I can tell you one thing. In this testing I've changed settings in my PC NIC like Duplex and Speed, LSO, QoS, etc and even though everything is back where it was the NIC drops connectivity after about 5 min now. I just can't get a break lol. Ill try reloading drivers for it when I get home. Never had any trouble with connection before making config changes. Also, when I make any NIC changes it locks up the config window and requires me to reboot the computer to get connection again.

 

No trees were harmed in the sending of this message, however, a significant number of electrons were terribly inconvenienced.

 

 

 

 

Share this post


Link to post

Ok I think I got my PC NIC issue resolved which was caused by the testing.  Don't know what screwed it up, but uninstalling it in Device Manager and rebooting seems to have fixed the problem.  I am back to being able to transfer 110 MB/s to a windows PC on the same switch as the Tower.  For fun I unplugged the network cable from the Tower and Laptop and switched them.  No Change.  Still 110MB/s to the Windows PC, and 40-50 MB/s to the UnRaid box.  I am using the same test file. 

 

I really appreciate any help you guys offer.  This is beyond my ability.  @johnnie.black I saw an older post from you about enabling direct IO mode.  When my parity check is finished I will stop the array and enable that mode.  Any chance that can help?

 

My server has always had these transfer characteristics.  I always wondered why I couldn't benefit from my gigabit connection and never realized that wasn't normal.  Now that I have verified I can when transferring to other machines I am REALLY wanting to fix the problem.  Is there any way the AMD based motherboard can be the issues?  I suppose I can try another PCI slot.  I really don't want to use my 2nd PCIe slot because I will need another controller at some point. 

 

Attaching diagnostic log.  Any advice is greatly appreciated. 

tower-diagnostics-20170402-1426.zip

Share this post


Link to post
39 minutes ago, allischalmersman said:

I saw an older post from you about enabling direct IO mode.  When my parity check is finished I will stop the array and enable that mode.  Any chance that can help?

 

Doubtful, it helps reads/writes with 10GbE, sometimes helps read speed with gigabit.

Share this post


Link to post

Do you have anything else installed on PCI slots? I'm not talking about PCI Express slots, just the old legacy PCI slots.

Share this post


Link to post
16 minutes ago, BRiT said:

Do you have anything else installed on PCI slots? I'm not talking about PCI Express slots, just the old legacy PCI slots.

No.  Only the NIC which has been changed out yesterday for another NIC

Share this post


Link to post

This doesn't make sense,  unless the PCI slot is sharing bandwidth with another device(s) or some other limits are in place. Maybe some odd sort of motherboard design.

Edited by BRiT

Share this post


Link to post

What I would now be looking is for the cause of the dropped packets in this section as shown in ifconfig.txt (in the config folder) in the diagnostics file.  My server has downloaded over 100M packets ( ~87.7 GiB bytes.) with zero drops!    

eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
        inet 192.168.1.2  netmask 255.255.255.0  broadcast 0.0.0.0
        ether 00:XX:23:XX:68:XX  txqueuelen 1000  (Ethernet)
        RX packets 80264801  bytes 102072657029 (95.0 GiB)
        RX errors 0  dropped 123649  overruns 0  frame 0
        TX packets 97820524  bytes 125457323456 (116.8 GiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

 

If you have Tips and Tweaks installed, you might want to play around with some of those NIC 'tuning' parameters to see if you can get rid of these. 

Share this post


Link to post

Yea before reboot I was over 2 million packets dropped.  From what I see on this forum it is normal for these Intel Pro 1000 NICs but these cards are pretty highly recommended.  I have tried 2 single port Pro 1000s and a dual port Pro1000.  All have exhibited this high packet drop.  I could try another PCI slot.  I will tinker with some settings.  Is there anything people do to run these Intel Pro 1000 cards other than plug them in?  I didnt load any drivers or anything.

Edited by allischalmersman

Share this post


Link to post
7 minutes ago, garycase said:

What SMB version are you using?

 

Sorry to ask such a newb question but how do you tell?  I ran smbstatus via telnet and I get Samba version 4.5.7  is that it?

 

Edit: Unter the user nobody and my IP of my desktop I am showing SMB3

smb.JPG

Edited by allischalmersman

Share this post


Link to post

Try adding these lines to the "extra configuration" section for SMB (on the Settings page)

 

max protocol = SMB2_02
allocation roundup size = 4096

 

Share this post


Link to post

... Not sure, but I think you have to reboot the server for that change to take effect.

 

Share this post


Link to post
5 minutes ago, garycase said:

... Not sure, but I think you have to reboot the server for that change to take effect.

 

Didn't have any effect even though the setting appears to have "stuck".  I will try a reboot after the parity check is complete (1.5TB left).  This is all I had to do correct?

smb extra.JPG

Share this post


Link to post

I believe that's correct.    I don't think the protocol actually changes until a reboot -- but as I noted above I'm not certain of that.   But I do know that some folks have experienced slower transfers when using SMB v3 (I think this is primarily with Windows 10 clients).

 

I'd reboot after your parity check and see if anything changes.

 

Share this post


Link to post
Just now, garycase said:

I believe that's correct.    I don't think the protocol actually changes until a reboot -- but as I noted above I'm not certain of that.   But I do know that some folks have experienced slower transfers when using SMB v3 (I think this is primarily with Windows 10 clients).

 

I'd reboot after your parity check and see if anything changes.

 

Well I will tell you both PC's used in testing (a x-99 platform desktop with pretty hot hardware, and a surface pro 3) are both W10 clients so that sounds hopeful. Those 110 MB/s transfers were between the two W10 PCs.  I guess I can bring my laptop home from work tomorrow to test.  It's a win7 PC. 

Share this post


Link to post
25 minutes ago, allischalmersman said:

Yea before reboot I was over 2 million packets dropped.  From what I see on this forum it is normal for these Intel Pro 1000 NICs but these cards are pretty highly recommended.  I have tried 2 single port Pro 1000s and a dual port Pro1000.  All have exhibited this high packet drop.  I could try another PCI slot.  I will tinker with some settings.  Is there anything people do to run these Intel Pro 1000 cards other than plug them in?  I didnt load any drivers or anything.

 

 

It has been several months (and, in those days, my systems was AMD MB's with Intel NIC cards --which were needed because of other issues) since I was working on a download issue to my PC's that I was playing around with those parameters and I seem to recall that changing some of those Tips and Tweaks NIC 'tunables' would affect those types of errors.  That is why I suggested playing with them...

Share this post


Link to post

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.