dslreports logo
    All Forums Hot Topics Gallery


Search Topic:
share rss forum feed


Kitchener, ON
reply to Teddy Boom

Re: [Cable] Thomson Thomson dcm425

@Teddy Boom: I do have to agree that you are technically correct to a certain degree from my experience. If you're connecting two Windows computers together, or if you're using a D-Link router product. Depending on your configuration and whether you're using a Linux based (such as Linksys or TP-Link) home router, or connect a Linux "server" directly to another computer, you can yield a much different result.

For example, when Bell started getting really bad for trying to throttle certain traffic, I got a little creative to get around this problem. I took my Simmens SpeedStream 5200 and put it into bridge mode, then connected it directly to /dev/eth0 on the Linux server. I assigned that network card a 10.0.0.x IP for the dummy purposes of binding an address to it for DSL. For cable, I configured it to run the internet connection to the cable modem, since I had to test it with a Rogers connection when I saw the results of Bell. I then connected /dev/eth1 directly to a Windows machine (ran the tests with Windows XP, Windows Vista, and Windows 7), assigning it a 192.168.1.x IP. From there, I set the Linux box to use /dev/eth0 to create my ppp0 link and set the standard firewall to deny all connections either incoming or outgoing, except for what I manually specified was allowed. This ensured that I would have no interference with the Windows packet scheduler trying to cram its proprietary Micro$oft crap down my throat, and without Bell's little "tracer" packets being received successfully. More or less, the Linux system was in "stealth" mode, which served as quite the effective firewall.

What was more interesting when I did this was that I noticed my internet speeds seemed to suddenly increase, and I got a much more consistent reliability out of it. Even stranger than that was that once I started using this with a backbone network, I noticed that my connection speeds between the Linux system and any other system were actually faster as well. I played around with a few other configurations (as to be able to better understand what configuration is best for customers) and noted that there were some interesting differences when D-Link products were avoided, and when a Linux system was introduced. Strangely enough, even a Linksys WRT54-G will actually perform faster when a static IP is assigned to its WAN port and no cable is connected. When not using the WAN port to tunnel internet traffic through and all routes to WAN were disabled to be able to again bridge the router through a Linux box, the difference in speeds both for internal networking and for internet connection (by tunnelling through the Linux box instead of the router's WAN) seemed to be faster and more consistent.

Teddy Boom
k kudos Received
Toronto, ON
Interesting.. I know windows does a lot of stuff that it really shouldn't be doing...
For example, when I go from one device to another to another, say to check MAC address in cable modem firmware on 10 boxes in a row, Windows web browsers will lock right up for 2-3 minutes. Makes no sense.

Anyway, the thing is, there don't seem to be a lot of recommended speed tweaks for windows networking. Or, there weren't a few years ago when I would have been all over that stuff if it was around. So, your experience is somewhat surprising.
Expand your moderator at work