Network bandwidth/transfer rate question (internet transfers faster than intranet!)

Ledneh

I've been trying to diagnose for a while why file transfers between my PC and my NAS are slow--often less than 1mbyte/sec over a wireless N network (or hell, even over a wired gigabit network). While doing various and sundry tests, I discovered something odd: no matter how I connected to my router, I could download from the Internet far faster than I could transfer files between machines on my local network, be it PC to PC, PC to NAS, or NAS to PC. In fact I could fully saturate my Internet downstream of 30 mbits/sec if I found a good site (Steam, for example).

I tried running iperf from several of my machines to my router (and vice versa; the router has DD-WRT installed), and came up with odd results: while directly connected to the router I could get the full gigabit bandwidth, but transfers from other PCs or the NAS were STILL below 1 MByte/s. Furthermore, while running iperf over wireless I got an average of 500 kbytes/sec bandwidth--and of course as soon as I stopped the iperf run and started an Internet download I could get six times that. Ditto when I tried testing over a powerline connection (though slightly faster in the file transfer department, about 750kbytes/sec).

I suck at networking and running iperf is the most complicated thing I've ever tried to do, so I'm out of ideas. What could be the cause of this oddball behavior? What other troubleshooting steps can I take? Why is networking such a god awful profession?

punk

What does your network topology look like? What is the make/model your network hardware? Also, if I knew what machines were connected to your network and via what medium, that'll help build the mental picture of your network.

If you really want to go nuts, you could run a Wireshark capture from your PC; one for PC to Internet, one for PC to PC, one for PC to NAS, and send them to me. I can probably pick out the cause from that and it might be faster than dickering back and forth here. Just don't log into any unsecured sites when you do it, I don't want to receive usernames/passwords - full disclosure: Wireshark will capture every packet to and fro, so I will see ANYTHING in plaintext you send/receive.

If you're interested in the Wireshark option, I can provide more detailed instructions.

Ledneh

I may take a look at Wireshark later, yeah, thanks for the offer. As for network topology, not much special:

- The Internet
-- Motorola SurfBoard cable modem of some sort
--- Asus RT-N16 router (gigabit ethernet 4-port switch, 802.11N) with DD-WRT installed so I can run iperf (didn't change network performance from stock at all)
---- DNS-321 NAS (gigabit ethernet)
---- PC 1 (gigabit ethernet)
---- PC 2 (802.11N OR gigabit ethernet, either directly or via Powerline)

Like I said earlier, PC-to-PC or PC-to-NAS or NAS-to-PC copies (and iperf runs) are hideously slow compared to Internet downloads, whatever those speeds might actually be (with direct ethernet being fastest and 802.11N being slowest, and Powerline generally being close to direct ethernet).

(also I didn't specify the make/model of the NICs in the PCs because I genuinely don't remember, sorry. PC 2 is a Lenovo laptop from 2 years ago, and PC1 is built into the Gigabyte motherboard from 3 years ago, but that's all I recall)

punk

OK, so you do have an integrated switch. That's what I was curious about. It makes me think the RT-N16 is having a Layer 2 to Layer 2 switching issue. From PC to Internet you're going from Layer 2 to Layer 3 and back, so it would switch those packets a little different. It's kind of hard to tell without being intimately familiar with that router, its hardware and the DD-WRT firmware.

We'll try some more basics first:

1) If you reboot the router, does the problem persist?
2) Are you running the latest version of DD-WRT? If not, have you considered upgrading? If you are, have you considered downgrading?
3) Are both PCs are able to pull 30Mbps (ish)?

Wireshark is really an incredible tool. An example: I have a Linksys E4200 that ran into some DHCP trouble a while back after I had updated its firmware. Because if Wireshark, I was able to quickly ascertain that the router was not responding to my DHCP broadcasts. Really cuts down on the troubleshooting if you know what you're doing with it.

Ledneh

Oh yeah, all those basic questions are covered: I've rebooted the whole thing god knows how many times, I've tried all sorts of firmware versions (various versions of DD-WRT, Tomato, Tomato-usb, stock), and all the PCs can pull full speed from the Internet on Ethernet (slower on other connections, obviously)

I'll be getting Wireshark this weekend and poking about if I don't figure it out by then--I don't have enough time for more than quick troubleshooting steps in the meantime.

If it's possible the switch part of the RT-N16 is goofed up, would it be possible to buy another switch from Frys, turn off the switching in the RT-N16, and run the network as Internet->Modem->RT-N16->new switch->PCs+NAS? Or does it not work that way? I know dick all about networking

punk

Yup, that would work. With all of your machines connected to the secondary switch, and the secondary switch connected to the RT-N16, that would be a quick way to determine if it's really a L2 to L2 issue since your L2 to L2 traffic will be isolated to that new switch. I'd be really curious to see how switch1-L2 to switch2-L2 traffic behaved, if that is indeed th eproblem.

You can pick up a decent little Linksys/Cisco switch for $25-30.

I don't suppose DD-WRT has any built-in diagnostic tools? Something that would let you look at statistics, error counters, etc.? It infuriates me to troubleshoot home network gear because I'm used to enterprise-level hardware that I can pick apart at incredible detail to find the problem. I troubleshoot home/SOHO hardware with a hammer.