Copying file from A-->B kills internet and network but B-->A is fine

[TanyaC]

Some time ago I had a problem with my Windows server 2012 R2 downloads failing. My ISP of course blamed my equipment. I had an ASUS RT-AC878U. So I replaced it with an RT-AC88U. That didn't fix the issue. It turned out to be a dodgy port on my Fibre Terminating Device ("NTD").

When I purchased the new router I also replaced all patch cables with new CAT6 leads. I also removed a Netgear Smart Switch GS108T-V2, as the RT-AC88U has 8 gigabit ports. This really was about eliminating a point of failure. I subsequently sold that switch.

I do a lot of video work on My PC. Once done I store the files on my server. My PC is "A" and the Server is "B".

Copying files to A to B has little to no impact on the network or the internet. However, when I copy from B to A everything stops. The Internet becomes unusable (Web pages won't load, games disconnect and lag really bad, pings go from 100ms to 3000ms and repeatedly get "Request timed out", speeds tests drop from 100mbps to 1mbps). Also, browsing the network becomes unresponsive, anyone watching content from the server will experience freezing, stuttering or PotPlayer or VLC going into "Not Responding".

I'm wondering if I did the wrong thing by routing all patch points through the router. Previous the RT-AC87U had one Wan connection, and a LAN port connection to a switch, and all patch panel points connected to that switch.

Can I please get some advice as to how I should structure this network?

Should I purchase another switch and put things back how they were?

To give you some background, there are (Over 2 floors) 10 PCs, 3 TVs, 2 Blu-ray players, 5 cell phones, 2 tablets and 2 laptops. For gaming there are 2 PS3s a PS4 and a Wii. I also have a Cisco SPA112 ATA for my VOIP service. All of this equipment is connected via 4 patch panel points. Each switch used is a Netgear GS108T-V2 except one DLINK DGS108D a port switch downstairs where a HTPC, TV, Blu-ray, 1 PS3 and a Wii are connected. There are 4 switches total

 

[System Error Message]

The netgear prosafe line is pretty good, shame you sold it. Its really simple, you're maxing out your 1Gb/s port that your PC doesnt have enough LAN bandwidth to use for internet too.

 

[TanyaC]

Thanks for your reply.
If you are saying it's better to route all devices through the switch with only one cable to the router, I can easily grab another switch. They are not that expensive.

But am I missing something - why is this a problem in one direction only?

And what's even weirder - This only started a few days ago. I've been running the same software, over the same infrastructure for many, many years and never seen this. I thought maybe "Blame Windows 10", but I've been running 1703 since it was released.

Could this be a failing switch?

However, I've been doing some more experimentation since my last post and I fear it may not [only] be an infrastructure problem. I was downloading the latest Linux mint distros via uTorrent. When uTorrent is running and I copy a file FROM the server, the problem is present. So I tried with a couple of other downloads like big buck bunny and some public domain documentaries. Sure enough, the problem is always present when uTorrent is running AND I copy a file FROM the server, but at no other time.

 

[FatherLandDescendant]

when uTorrent is running AND I copy a file FROM the server, but at no other time.

Use a different torrent program and try your test again. I recently found that uTorrent was making my VPN crash in my pfSense box, the traffic from uTorrent was running up against a policy rule in Snort after it was running for a bit and the VPN would shut down. Once I figured out that Snort was causing the issue and put in a suppression rule for uTorrent it hasn't crashed since.

 

[TanyaC]

Ok, I'll do that. I've tested pretty much all torrent programs when I was testing Windows 10 for deployment to the PCs here, and nothing really comes even close performance wise. I haven't actually updated uTorrent since 3.4.7.42330 and all was well until a few days ago... I will do a clean install of it as well.

Will let you know how that goes.

 

[TanyaC]

Ok, problem solved!

I did try different programs, but the result was the same. And the the two operations (downloading and copying a file), did not have to be on the same PC. If I was running uTorrent downloading Mint from A, and copying a file from B --> C (B being the server so it was a "pull"), the problem surfaced.

Anyway; The solution - Put a switch back into the loop as it was previously. So now, the router goes to the switch and the switch is connected to all patch panel ports and feeds the rest of the house.

I learned something new today - never connect all devices directly to the router.

 

[TanyaC]

I originally asked this question on another forum. I finally got a reply when I posted that I had solved the problem. This is what the networking person said...

I suspect it is just a coincidence that it fixed it or you had some other issue.

If you where to open most routers up you would find a small switch chip running the lan ports. This in many cases was the same chip used by say a small 5 port external switch. What you have in effect is a switch with a single cable going to the router chip leaving 4 ports available to you for end devices.

The switch chip in general can run all ports at maximum speed up and down at the same time. So with 4 gigabit ports a total of 8gbit of traffic could pass at the same time....not that there is a realistic configuration that would use that.

Hard to say why you fixed it but it is not because you use a external switch.

Is he right?

 

[FatherLandDescendant]

Is he right?

Pretty much. The LAN ports on a router are just a switch, one that is integrated to the routers board, but just a switch all the same.

 

[TanyaC]

Ok, I've found other people who can also reproduce the problem. By removing the switch between their router and the devices they too lose Internet connectivity when copying files unidirectionally. So we have ASUS, Netgear and D-Link routers tested, and all exhibit the same symptoms.

I'm happy to accept that it's not the router that's gagging under the load, so where can I start looking for the "real" cause?

Sadly, my post on other forums now has me labelled as an anarchist and spreader of false information. All I want to do is solve the problem. As I was able to reproduce the issues consistently with multiple switches (Netgear and D-Link), across all PCs on the network, I honestly figured I had a solution.

Remember, This happens across multiple PCs on the network. If any of half a dozen torrent programs are actively running, any copy from the server to any PC on the entire network causes the internet to drop and the PCs to become unresponsive.

Grabbing at straws here - Are the switches in routers intelligent, or do they broadcast all traffic across the whole network? Whereas the Netgear switches are "smart"

 

[TanyaC]

Here is my network.
HOME-LAN-1.jpg is the configuration that works
HOME-LAN-2.jpg is the configuration in which the problem presents.
The only difference is the lower left corner in the NBN Cabinet. -1 has a switch, -2 does not.

So what's wrong with the network design?

https://ibb.co/c8Tk1w

https://ibb.co/eoUfZG

 

[degrub]

sounds like the router CPU is getting involved for some reason because of the type of traffic.
Might ask RMERLIN about the issue since he develops alternative firmware for routers.

 

[TanyaC]

I am running the latest Merlin Firmware.
I'll ask on that forum.

 

[System Error Message]

Whats happening is that traffic is going to the CPU of the router. On a typical consumer router, there are 2 1Gb/s links from the CPU. 1 to WAN and 1 to switch. So if the link to CPU is full then the you get a bottleneck. Usually this only happens if you bridge through CPU rather than switch. Seeing as any LAN port can be used as a 2nd WAN port, it may be bridging instead of switching and having traffic go through the CPU instead.

The CPU is plenty fast for bridging.

 

[TanyaC]

Whats happening is that traffic is going to the CPU of the router. On a typical consumer router, there are 2 1Gb/s links from the CPU. 1 to WAN and 1 to switch. So if the link to CPU is full then the you get a bottleneck. Usually this only happens if you bridge through CPU rather than switch. Seeing as any LAN port can be used as a 2nd WAN port, it may be bridging instead of switching and having traffic go through the CPU instead.

The CPU is plenty fast for bridging.

I don't think I'm following. You say that if it's bridging through the CPU then I will get a bottleneck, but then say the CPU is plenty fast enough for bridging???

I assume the bridging through the CPU is the default behavior?
I also assume that the addition of a switch takes that load off the routers CPU completely, since the problem is not present when the switch is present?
I assume when you refer to switch you are referring to the routers switch, not the Netgear GS108T-V2.

So.. Is the switch "Hiding" a different problem, or solving my problem?
If bridging via the CPU is causing a bottleneck, how does one change the router to bridge through the switch. Is that a recommended practice, and are there any downsides to doing that?

 

[degrub]

let the router "route" and the switch "switch". if there is any CPU involvement in the router switch, then the routing will bog down under your loads. If the router switch is not involved in the switch activity - either through active processing of packets or shared buffer memory, then you should not see the slowdown.

The use of the independent switch to "solve" the issue suggests that is what is going on in the router. RMERLIN may be able to confirm since you are using that software.

 

[System Error Message]

I don't think I'm following. You say that if it's bridging through the CPU then I will get a bottleneck, but then say the CPU is plenty fast enough for bridging???

I assume the bridging through the CPU is the default behavior?
I also assume that the addition of a switch takes that load off the routers CPU completely, since the problem is not present when the switch is present?
I assume when you refer to switch you are referring to the routers switch, not the Netgear GS108T-V2.

So.. Is the switch "Hiding" a different problem, or solving my problem?
If bridging via the CPU is causing a bottleneck, how does one change the router to bridge through the switch. Is that a recommended practice, and are there any downsides to doing that?

its not the CPU thats the bottleneck, its the link between the CPU and internal switch chip thats the bottleneck. Typically this shouldnt be the case with asus router's broadcoms chipset having the switch integrated into the CPU but you've proven that this is the case. Shame on you broadcom.

 

[FatherLandDescendant]

or solving my problem?

If I'm following System Error Messages' line of reasoning correctly the external switch is solving your problem by assuming the burden of processing the file transfer, leaving the router free to just be a router. The supposition is that the router, pulling double duty, is deciding under load that internal network traffic management ranks of higher importance than external internet network availability.

When you recreate this issue, and loose internet connectivity then when the file transfer job is complete, does you internet connectivity return without intervention on your part?

 

[TanyaC]

When you recreate this issue, and lose internet connectivity then when the file transfer job is complete, does you internet connectivity return without intervention on your part?

Yes.

So, there is no problem that the extra switch is hiding and I'd be wasting everyone's time trying to find and solve it?

As in - this is a router issue so (1) adding the extra switch back was the correct solution and (b) plugging LAN segments into the router is not a good thing as it "causes" this problem.

The person on the other forum has been beating me up telling me I'm spreading misinformation was the one who was wrong.

If that's true - I'm sorry to have wasted everyone's time!

 

[TanyaC]

Ok, this appears to be the issue and solution....
https://www.snbforums.com/threads/c...kills-internet-and-network.44192/#post-375667

 

[coxhaus]

I agree with letting the router route and the switch, switch. I have ben saying always run a separate switch with a router. Don't use a lot of little switches either. Use one large switch for all connections since the bigger back plane makes the difference in performance. If you want real high speed use a layer 3 switch.