Download at 1Gbps, upload at 100Mbps

[kriegalex]

Hi,

I've been experiencing a weird thing that I wasn't able to find online, at least not easily.

I recently switched from using powerline adapters (devolo) to running a CAT6 cable through the walls to my FreeNAS box to enjoy 1Gbps connection from my main computer (I will later connect all the rooms to a single 8 port switch).

What happened is that I naturally wanted to see that beautiful 100MB/s line on Windows when copying stuff through the network drives and I got it when downloading a file from the server, but when I upload something I only get 10MB/S (100Mbps).

What is even more strange is that when I go to my network adapter properties on Windows, it shows 1Gbps FULL DUPLEX and on FreeNAS it shows "Ethernet autoselect (1000baseT <full-duplex>)".

I don't know why, but when I go to my Intel adapter properties (Windows), you get all those nice Intel NIC features and after I press on the diagnostic buttons, something clicks somewhere on the network because after that I get 1Gbps upload, even though the diagnostic says "ok it works" (of course because my network works just fine, just the upload acts weird).

Do you recognize this problem ? How can I solve it so I don't have to do this diagnostic tool each time ? The machines are linked through a Netgear GS105eV2.

 

[anodos]

Hi,

I've been experiencing a weird thing that I wasn't able to find online, at least not easily.

I recently switched from using powerline adapters (devolo) to running a CAT6 cable through the walls to my FreeNAS box to enjoy 1Gbps connection from my main computer (I will later connect all the rooms to a single 8 port switch).

What happened is that I naturally wanted to see that beautiful 100MB/s line on Windows when copying stuff through the network drives and I got it when downloading a file from the server, but when I upload something I only get 10MB/S (100Mbps).

What is even more strange is that when I go to my network adapter properties on Windows, it shows 1Gbps FULL DUPLEX and on FreeNAS it shows "Ethernet autoselect (1000baseT <full-duplex>)".

I don't know why, but when I go to my Intel adapter properties (Windows), you get all those nice Intel NIC features and after I press on the diagnostic buttons, something clicks somewhere on the network because after that I get 1Gbps upload, even though the diagnostic says "ok it works" (of course because my network works just fine, just the upload acts weird).

Do you recognize this problem ? How can I solve it so I don't have to do this diagnostic tool each time ? The machines are linked through a Netgear GS105eV2.

Does iperf get close to gigabit speeds both ways?

 

[SweetAndLow]

Iperf is a great test. This sounds like a bad cable to me.

 

[jgreco]

Does iperf get close to gigabit speeds both ways?

I would definitely try iperf on each end. For Windows, make sure you're using iperf2, not iperf3. When you're seeing the 10MBytes/sec behaviour, quickly stop everything and run an iperf test in each direction, and let us know what the speeds are.

 

[kriegalex]

Windows server, freenas client :

[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 1.08 GBytes 930 Mbits/sec 0 sender
[ 4] 0.00-10.00 sec 1.08 GBytes 930 Mbits/sec receiver

FreeNAS server, Windows client :

[ ID] Interval Transfer Bandwidth
[ 4] 0.00-10.00 sec 1.08 GBytes 929 Mbits/sec sender
[ 4] 0.00-10.00 sec 1.08 GBytes 929 Mbits/sec receiver

With uploading at 11MB/s :

And downloading at 112 MB/s :

 

[jgreco]

Okay, but are you currently experiencing the slow transfer issue?

 

[kriegalex]

Yes, I provided the screenshot at 11.4MB/s instead of ~110MB/s when copying from Desktop to some folder on remote network drive.

EDIT : I rebooted the computer and now it is ok at 110MB/s upload. It's very weird. I will be waiting to encounter again 11MB/s and will do right after iperf tool

 

[anodos]

Did you do the wiring yourself? If not, did the installer certify the cables?

 

[jgreco]

Yes, I provided the screenshot at 11.4MB/s instead of ~110MB/s when copying from Desktop to some folder on remote network drive.

Which you edited in after my response. But fine.

Okay, that's strange. Normally what I'd say is that the transfer starts out with a burst and then a dip followed by flat performance which kind of suggests that things are running full tilt.

But a mere 10MB/sec and the fact that that's so close to 100Mbps speeds suggests networking issues.

So tackle this from both ends. Pull up the FreeNAS console and run "gstat" when this is going on. If it says that your disks are mostly running near 100%, you may have a problem with one of your disks, or something else going on that just happens to be causing it to be very-close-to-100Mbps speeds.

Then give us a rundown on your network. Is your home NAT gateway ("router") maybe a 100Mbps device? What sort of switching is there? Have you tested the cabling?

You ran Cat6 yourself? Check the system log in /var/log/messages to see if maybe it is renegotiating back and forth. Did you run solid wire? Did you try any Dumb Cabling Tricks like crimping RJ45's onto solid core wire?

 

[kriegalex]

Did you do the wiring yourself? If not, did the installer certify the cables?

Yes I did the wiring myself, following the CAT6 T568A pattern shown on the connector on both ends (and the Intel diagnostic seems to tell me the wiring is ok, otherwise it shows an error normally).

 

[kriegalex]

I will be waiting to encounter again 11MB/s and will do right after iperf tool and all the things you suggested. My switch is a Netgear GS105EV2 fully 1Gbps capable and the fact that I'm getting right now 110MB/s proves it. I just don't understand why sometimes it goes down to 11MB/s.

Your theory of disk usage is interesting and I didn't think of that, I will try to see if disk usage is correlated to 11MB/s upload. This would imply disks can't write at more than 11MB/s if something else is going on, while they can read at 110MB/s which seems logical, because they have more read performance than write (right?).

 

[jgreco]

I will be waiting to encounter again 11MB/s and will do right after iperf tool and all the things you suggested. My switch is a Netgear GS105EV2 fully 1Gbps capable and the fact that I'm getting right now 110MB/s proves it. I just don't understand why sometimes it goes down to 11MB/s.

Your theory of disk usage is interesting and I didn't think of that, I will try to see if disk usage is correlated to 11MB/s upload. This would imply disks can't write at more than 11MB/s if something else is going on, while they can read at 110MB/s which seems logical, because they have more read performance than write (right?).

Ehhhhhh. Hard to know. Actually write usage tends to dominate in my experience, but we need more clues here.

 

[kriegalex]

I randomly got a 11MB/s copy to network drives and iperf DID show 100Mbps perfs while the copy was running :

PS E:\My Files\Programs\Iperf\iperf-3.1.3-win64> .\iperf3.exe -c 192.168.1.150 -p 5201
Connecting to host 192.168.1.150, port 5201
[ 4] local 192.168.1.85 port 51494 connected to 192.168.1.150 port 5201
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-1.00 sec 11.5 MBytes 96.5 Mbits/sec
[ 4] 1.00-2.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 2.00-3.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 3.00-4.00 sec 11.6 MBytes 97.6 Mbits/sec
[ 4] 4.00-5.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 5.00-6.00 sec 11.5 MBytes 96.5 Mbits/sec
[ 4] 6.00-7.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 7.00-8.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 8.00-9.00 sec 11.5 MBytes 96.5 Mbits/sec
[ 4] 9.00-10.00 sec 11.6 MBytes 97.6 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-10.00 sec 116 MBytes 97.2 Mbits/sec sender
[ 4] 0.00-10.00 sec 116 MBytes 97.1 Mbits/sec receiver

Right after the copy I relaunched iperf and it was at 1Gbps again and the copy to network drives also. It's like random I don't get it. The cable can't be faulty 10 sec and the next 500 sec work just fine.

For the drive usage theory, it's not that, I launched an internal copy from disk to disk to stress the disks at ~80-90% and the copy was fine at 110MB/s (I always talk about upload since it's the problem).

I also looked at gstat while the 11MB/s copy was running and the disks were near idle, with slight peaks at 30% usage.

EDIT : I will try to run another cable (not through the wall) and see if I can see the 11MB/s thing to know if I have to investigate my cable

 

[anodos]

I randomly got a 11MB/s copy to network drives and iperf DID show 100Mbps perfs while the copy was running :

PS E:\My Files\Programs\Iperf\iperf-3.1.3-win64> .\iperf3.exe -c 192.168.1.150 -p 5201
Connecting to host 192.168.1.150, port 5201
[ 4] local 192.168.1.85 port 51494 connected to 192.168.1.150 port 5201
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-1.00 sec 11.5 MBytes 96.5 Mbits/sec
[ 4] 1.00-2.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 2.00-3.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 3.00-4.00 sec 11.6 MBytes 97.6 Mbits/sec
[ 4] 4.00-5.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 5.00-6.00 sec 11.5 MBytes 96.5 Mbits/sec
[ 4] 6.00-7.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 7.00-8.00 sec 11.6 MBytes 97.5 Mbits/sec
[ 4] 8.00-9.00 sec 11.5 MBytes 96.5 Mbits/sec
[ 4] 9.00-10.00 sec 11.6 MBytes 97.6 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-10.00 sec 116 MBytes 97.2 Mbits/sec sender
[ 4] 0.00-10.00 sec 116 MBytes 97.1 Mbits/sec receiver

Right after the copy I relaunched iperf and it was at 1Gbps again and the copy to network drives also. It's like random I don't get it. The cable can't be faulty 10 sec and the next 500 sec work just fine.

For the drive usage theory, it's not that, I launched an internal copy from disk to disk to stress the disks at ~80-90% and the copy was fine at 110MB/s (I always talk about upload since it's the problem).

I also looked at gstat while the 11MB/s copy was running and the disks were near idle, with slight peaks at 30% usage.

EDIT : I will try to run another cable (not through the wall) and see if I can see the 11MB/s thing to know if I have to investigate my cable

Sounds like a good plan. A few things to look out for when running cable yourself:

• Use right type of cable (may be worth looking at your local building codes). Typically you'll want solid-core CMR.
• Don't run it through A/C ductwork
• Don't run network cable in the same conduit as your electrical wiring (or too close to them).
• Avoid kinking cables
• Don't exceed 25 ft/lbs of pulling tension when running the cable. (In other words, try to feed the cable smoothly through the wall or conduit - no jerking).
• Avoid having overly sharp bends in cable.
• Don't run cables too close to your HVAC system, attic fan, etc.
• Untwist the twisted bare the bare minimum necessary to properly terminate the cable in the jack.

 

[SweetAndLow]

This sounds like bad cabling that gets wiggled and works sometimes. Can you force 1000mb/s connection and disable auto negotiation? After doing this do you get a working connection? If not it could be wire or port on your switch. Are you using any link aggregation?

 

[zoomzoom]

Depending on how you ran the cable through the walls, you may want to grab a multimeter and check resistance.

• If the in wall cabling terminates in a non-OEM jack connection, I would ensure the leads on either end are properly seated, with no fragments of insulation left on the bare leads. If in wall cabling terminates in an RJ45 jack you crimped yourself, I'd recrimp the end since it's quick and easy.

If you ran the cable yourself, did you verify it's not running alongside wires that could be creating EMI? There's only a few reasons why an ethernet cable would be full speed in one direction but have intermittent issues on the other 2 wires (Tx).

 

[kriegalex]

I used the RJ45 end showed in the image, following the pattern A shown on the rj45 plug. The cable is running inside a standard tube that contains nothing else. I don't know if other tubes are running next to the RJ45 one.

For the problem, it may indeed be the cable, I will try to redo the connection to the wall plug. On the pattern shown for 568A, what are the wires that handle upload that may be wrongly done so that I inspect them carefully ?

EDIT : Murphy's law being Murphy's law, as soon as I posted this answer, I got a 100Mbps upload with another "commercial" cable (received with the modem) that I just run on the floor (not through the wall) and with another computer.

So it seems it's not the cable and it's not the computer, so the last things remaining are the switch or the NAS I guess. I will investigate the switch first.

 

[zoomzoom]

Electrical cables are easy to test with a multimeter...

• 1st, test continuity to verify there's not an open circuit
  • This is done by selecting the continuity icon on the multimeter (the symbol is a dot with radiating waves from it)
    • Place each lead on either end of the specific wire being tested
  • If the wire tested is a continuous circuit, the multimeter should read 0 [ohms] and sound a continuous beep (this signifies a dead short, thereby demonstrating the wire is a continuous circuit)
• 2nd, test resistance
  • You will need to determine via google what the proper resistance [ohmage] is for the gauge and length of wire within the cat6 cable, as gauge can range from 22 - 36.
  • To determine the resistance, you must select the next highest ohmage value above what the proper ohmage should be for the cable on the multimeter.
    • Place each lead on either end of the specific wire being tested

Provided the above two demonstrate correct values, there's nothing wrong with the electrical cable.

 

[maglin]

So you shouldn't put RJ-45 ends on solid wire? I'm not a network guy but I did run a lot of CAT5e through my house and have used left over cable for patch cables. It used solid wire. I thought all CAT5e cable used solid wires. If that is wrong I'll need to replace several cables in my network. The wires in my walls are all punched down into jacks.


Sent from my iPhone using Tapatalk

 

[zoomzoom]

So you shouldn't put RJ-45 ends on solid wire?

Solid vs stranded shouldn't matter at such a small gauge. Generally speaking, stranded electrical wire is used when flexibility in the wire is required (i.e. automotive electrical wire vs solid electrical house wires).