Please help me understand my LAN / router / switch problem

[Bart]

Hi all,

Today, I stumbled onto the weirdest problem, and I have no answer for it.

This is my home network (only the wired parts, the wireless is not important here)
1 D-Link DIR-655 router (with 4 gigabit LAN ports)
1 HP Procurve 1400 24-port switch (all gigabit)
1 24 port patch panel
everywhere Cat 7 cable inside the walls
1 Windows 8 PC - hardware only 3 months old (gigabit)
1 unRAID server (gigabit)
1 laptop (gigabit)

The router is connected via one Cat6 cable to the Procurve 1400. The Procurve then serves the home for wired internet. For this, I need to click a patch cable into the switch and then into the patch panel. I needed to put this patch panel into place because I used Cat 7 cables inside my walls and I couldn't put RJ-45 connectors onto Cat7 cable, hence the instalment of a patch panel. (which turned out to be a good idea in hindsight because it allowed me to use my switch very flexible)

Now, today I was busy installing my first server, an unRAID server. After I had set it up, I started copying the files to it using TeraCopy. Immediately I noticed the slow transfer speeds of about 11 MB/s. I had worked with my network for two years now, but it's mainly used for providing internet access to all the drop points. No file copying (otherwise I would have noticed this sooner). And so the troubleshooting began.

I installed Iperf onto the unRAID server and onto my Windows 8 PC and ran the test. Result : 10.8 MB/s. Hmmmm...

I then started working differently. I found this very nice article by Tim which talks about 5 ways to test the network speed. I decided to start using LAN Speed Test Lite, for the sheer straightforwardness of it.

I hooked up my gigabit Asus Laptop (windows 7 - 1 year old) and ran the test via LAN Speed Test to a share on my laptop. The result: 12 MB/s.

Ok, now the gloves were off. I still had a small 8 port Neatgear switch laying around (Netgear GS-108), and decided to cut the Procurve from the equation and put the Netgear switch to use. I connected ONLY my PC and my laptop and re-ran the test. With the same result. So the Procurve was not at fault here.

Then I started fearing that the CAT7 cables inside my walls were somehow faulty. So I put up a very simple test setup, with factory-made patch cables : d-link router -> netgear switch -> PC & laptop (connected to Netgear switch). The result : the same!! So my wall cables were not at fault here.

Then I decided to also drop the Netgear switch and set up the simplest setup there is : D-Link router, with the PC and laptop connected to it's own internal LAN ports. The result : BAM! 56 MB/s write, 106 MB/s read!!

Even though the write speeds were still a bit low, the read speeds were ok. So now, this is my situation: for some reason, when I connect a switch, ANY switch, to the LAN ports of my D-link DIR-655 gigabit router, and then my devices to that switch, the speed of the network goes down to 100 base-T speeds. If I connect my devices directly to the LAN ports of the router, the speeds (at least the read speed) are normal.

What might be causing this? Is my router on it's way out? I've looked though it's settings and cannot find any setting which could cause this behaviour

I'm very much inclined to order a new router at this point, but I wanted to hear the feedback of this forum first.

Thank you all for your time and any help you may provide.

 

[thiggins]

Good troubleshooting! Sounds like auto-negotiation is failing between the DIR-655 and both of the switches you have. What do the switch lights show for link speed?

The router is the common factor, so probably the culprit. (Could also be the router's power supply.)

 

[Bart]

Thanks!

I've just looked at the LED of the Procurve (the Netgear isn't installed anymore), and it says Gigabite (which it clearly isn't)

I've found a temporary solution though. As said I'm using a patch panel. This is one of those cases where it comes in handy. I've now connected the cable of my main PC and the one of the server directly to the LAN side of the router. I've used a third port for connection to the Procurve switch for all other devices in the house. This should do fine for the moment and I've still got a fouth port left in case I'd like to connect a second server or something.

I will be looking at a new router though. Still a bit in doubt whether I'd buy a new AC model (even though I haven't got any AC devices yet), or just stick to a tried and tested N model which is more than likely enough for my purposes. (mainly browsing the internet with iPhone and iPad really).

I do want a router which is capable of running a VPN though (to be able to connect to my server from outside the house more easily and more securely), and I guess that will limit my options somewhat.

I'll have a look at the charts!

 

[Dreamslacker]

What happens if you do the following:

1. Connect only the desktop and laptop via patchcords directly to the switch without the router.
2. Set static IPs for each of the 2 devices (for the purpose of the test).
3. Run the test again after verifying that both devices are connected in GbE Full-duplex.

Seems like an extremely odd problem to have since the traffic shouldn't run through the uplink cable to the router (unless you had any of the hosts connected directly to the router).

 

[Bart]

Sorry for the tardy response, I've just haven't had the extra time to put into this problem.
Your suggestion is a good one though. I'll see if I can set this up asap and will report back here about it.

 

[ttran55]

I had a similar problem with my Netgear switch GS108 when using Cat6 cables. It works fine with Cat5e.