Teh Internet is broekn
April 25, 2006 7:55 AM   Subscribe

Please help me diagnose an intermittent computer network problem.

I have had some wonky network and internet behavior for the last few months at home and it is getting worse. I suspect there is a bad piece of hardware somewhere and I was hoping there might be software out there that could help me track down which one is the culprit. The frustrating part is that the problem is intermittent which makes it difficult to isolate. The Internet connection is sometimes down for a few minutes or a few hours. Even when it is working lately, there are lots of flickers. For example, I will try to open a page and it immediately comes back saying it can’t find it, I reload and it works but it is slow or some pictures never load. My novice guess is that packets are getting dropped somewhere, maybe predominately upstream ones. There may also be problems within the LAN (or maybe not). Occasionally when I am accessing a shared drive across the network I have gotten some strange behavior.

The problem is I don’t have enough equipment to really begin to swap out each piece of hardware individually to see if that solves the problem. Also, because the problems are intermittent it could take a long time. I was hoping there was maybe a program I could use to test how well each piece of the network was working. Any suggestions on a strategy?

I use a Comcast cable modem for my broadband. It is plugged into an old system running a Smoothwall firewall. That goes out to a 4-port D-Link gigabit switch, and that is hooked up to a 10/100 D-Link switch so I can connect everything. The systems hooked up to my network include two XP PCs, a ClarkConnect fileserver, an Xbox, and a network printer. So assuming the software is all OK, there are several pieces of hardware that could be the problem (cable modem, 2 NICs in the firewall, the switches, etc). Oh, and I have tried rebooting things with no apparent change in behavior.

Please help, this is driving me crazy!
posted by Tallguy to Computers & Internet (12 answers total)
 
Make sure every piece of equipment you have agreees about what time it is, then install Ethereal on everything in sight, then note the time the flakiness occurred and look for discrepancies in the Ethereal packet traces.

Or, you could try just unplugging and replugging every Ethernet cable connector a few times. That fixes the flakiness I sometimes get on my Dell laptop.
posted by flabdablet at 8:01 AM on April 25, 2006


In thei UK? Is this at home?

Are you using ADSL? - Do all your phone sockets have filters on?

Have you tried the PC somewhere where there's a known working network or a known working PC / laptop in your home network?

Sorry for so many questions, but you have to methodically reduce where the problem is.
posted by Dub at 8:34 AM on April 25, 2006


Dub: This is in the USA and it is at home. It is a cable modem. Everything worked fine until a few months ago, and since all of the computers in the network are desktops I have not had a chance to plug them into different networks.
posted by Tallguy at 9:31 AM on April 25, 2006


I'd start with plugging a desktop into the cable modem using one network cable, and it is persists, try another cable. If it's still bad, start with the cable modem, since you can generally exchange them at the cable company office for no cost and a short drive, plus they seem to go bad often in my experience.

If everything is affected, like it sounds, it's likely a cable or perhaps a piece of hardware nearest the cable connection.
posted by kcm at 9:43 AM on April 25, 2006


I will try to open a page and it immediately comes back saying it can’t find it, I reload and it works but it is slow or some pictures never load.

I had exactly that once; have a look here.
posted by ed\26h at 10:05 AM on April 25, 2006


ed\26h: Boy am I wary of running a strnage patch, but that actually seems to have made a difference. I will have to wait and see whether that cures all of my network's apparent ills.
posted by Tallguy at 11:02 AM on April 25, 2006


That patch should be required for all machines they've limited half-open connections to 10 on (post-SP2 XP?). I understand the meaning behind it, but boy was the technique implemented poorly.. I've found 50 or 100 to be more than enough without any detriment.
posted by kcm at 11:03 AM on April 25, 2006


XP had this before SP2 - I believe it was in the original release , and was one of the (many) reasons some people decided not to move to XP from 2000.

Just FYI (in case you didn't know), the 10 concurrent connections is put in place basically on purpose by Microsoft to prevent people running web servers on a workstation OS. In this day and age, 10 concurrent connections isn't unlikely nor are they saving us from ourselves in terms of technology. It should also be noted that it's 10 concurrent TCP/IP connections, which includes most of the stuff you're doing on the local network as well.
posted by hoborg at 7:43 PM on April 25, 2006


According to the site I linked (and a number of other reliable sources, check Google) the limit was introduced with XP SP2. Presumably if Microsoft wanted to prevent people from running web servers on their workstation operating systems they would not bundle their web server software in with them.
posted by ed\26h at 2:22 AM on April 26, 2006


We've hashed out this SP2 behavior before in AskMe, but:

Actually that change was put into SP2 to help prevent virused computers from thrashing a network. Many viruses try to propogate by making connections to random IP addresses - as fast as they can. Since the vast majority of connections are going to fail, MS put a limit of 10 concurrent half-open (ie, other side won't respond) connections per second. This prevents the infected computer from overhwelming your network, before it's caught and cleaned up.

You can tell if this is occuring during your errors by opening up a command window and typing netstat -b. Look at the list of http/tcp connections. Are many of them in a waiting state? If so, that may be your problem.
posted by Dunwitty at 3:42 AM on April 26, 2006


It should also be noted that it's 10 concurrent TCP/IP connections, which includes most of the stuff you're doing on the local network as well.
NO! It is 10 half-open connections. There is a MAJOR difference, and you are just spouting FUD.
posted by Rhomboid at 3:38 PM on April 26, 2006


Just in case anyone in the future trips over this thread. It turns out I continued to have the problem. Eventually, I found someone capabe on Comcast's customer service that mentioned he could set tp a job to continously ping the modem to determine if the problem was on their end. Turns out he got 20% packet loss and the problem was Comcast's. Techs came out and traced the problem to a bad cable fitting outside. Hopefully the problem is fixed now.

I have to give mad props to the Comcast guys I interacted with this time. They were professional and effective.
posted by Tallguy at 8:36 AM on May 3, 2006


« Older HTML woes: Google/Amazon ads and IE display woes   |   Problems with bittorrent files on OS X... Newer »
This thread is closed to new comments.