Network bandwidth troubleshooting

[DaveC426913]

I continue to have troubles with my residential DSL service. Occasionally (on the order of every half hour or more), my service will just stop working for a minute or so. (Gee, that doesn't sound like a big deal, but really, it makes me want to toss my computer out the window).

I'd like to find a comprehensive way of diagnosing the problem, say, a piece of software that will monitor traffic so I can see where it's hanging.

My network:

• Sympatico 3rd-party-resold DSL (through my employer) they've checked my sync rate, which seems fine
• old phone lines
• linkysys 802.11B 2.4GHz wireless access point - WEP enabled
• one computer on wired connection
• one laptop on wireless
• both have Norton Antivirus
• one has Skype and Limewire (checked these settings)

When I have a good connection, it's just fine. I've run net speed tests and they come back at 2Mb down and 500Kb up every time (or 0 and 0), So, there's no grey area in my pipeline, it's either wide open or blocked completely.


I have not been able to pin down a consistent set of conditions under which it occurs.
- Sometimes I can ping a server out there, even while my browser is dead, sometimes I can't.
- Sometimes it seems to affect Outlook, sometimes just my browser.
- Sometimes it's just my laptop, sometimes it's both.
- Sometimes it seems to be linked with my WebHost (which is on the other side of the world) and sometimes it even does it with Google.
- Sometimes 'Repair'ing my wireless connection clears it up, sometimes it doesn't.
- Sometimes a hard reboot of my modem and Access Point works sometimes it doesn't.
- It does NOT seem to matter whether someone is actually using the desktop or not - I still get the problems on the laptop.

(It is worth noting that, with a down time of only a minute or so, it's possible that none of these techniques is doing anything, that merely the connection returns after a minute or so, while I'm fiddling about.)

Oh woe!

 

[dduardo]

Some time ago my DSL would stop working and unplugging the power and reconnecting it would fix the problem. Eventually I got a new modem and it fixed the problem. I very rarely have downtime nowadays.

 

[-Job-]

Some DNS problem would cause all the symptoms you described. Two things you can do, if you haven't already, is during down time, do a tracert to some site, and ping by IP.
At the command prompt do:
tracert www.yahoo.com
It will display the round trip to each of the routers in the way. If it goes halfway but then stops, then there's something wrong with a server/router outside your LAN. If it doesn't even get to the first router, then it is a connection issue and you have to test to determine whether it is inside your LAN (bad router, bad link, interference...etc) or your ISP is the one at fault and you should really give them a call.
The most important thing is to determine whether it is within your network or outside it. Somethings you can do to determine this are:
- ping your router at down time, if you can't ping it, then the problem is inside your LAN.
- connect a machine directly to the modem at down time (unless you have a static IP) and see if everything works. If it doesn't then it's the modem/isp.
- leave a machine connected directly to the modem and see if connection goes down periodically as well.

I had a similar problem, not every half hour, but periodically (at most every hour) for a while, and it turned out it was my router that was not working correctly. I bought a Netgear router (my old one was from Linksys) and after that i haven't had any problems. It might be worthwhile to buy a new router to determine if it works correctly or not, and return it if it's the same thing.

 

[DaveC426913]

OK, I did a tracert and it timed out. I pinged (pang? pung?) and it timed out.

So I attempted to download the latest firmware for my Router. I have a BEFW11S4 v2.1 which, according to Linksys doesn't exist. They say it's got to be a 2.0. So, the firmware wouldn't take on my box.

Anyway, they figured, to get around the problem, they lowered my MTU and did some other router setting stuff.

No word yet on whether that's fixed the problem.

 

[DaveC426913]

BTW, how many hops is normal?

Is 12 hops and 600ms to Google OK?
Is 19 hops and 1500ms to my own webhost a little high?

Also, is is normal for the very first hop to take several seconds and then time out before continuing? I thought that had something to do with my having a wireless router, but it has the same effect when I bypass wireless access and use direct ethernet connection? Does that delay every connection I make or is that an artifact of the tracert command?

 

[BornLogician79]

just a couple of quick hardware ideas.. 2.4 giahertz phone in area cause same effect on my network. I've also installed the router and dsl modem into a cooling system due to overheating new out of box. The cooling did help the dropoffs. Turning the router upsidedown to expose vents to prevent bottled up heat on chipsets. maybee. depending on the design. Simular to cooling cpu, overheat results in lockup? my router is a hotbox, which did need the cooling. As soon as I pushed the power button to the cooling system ms would drop to game server. simple ideas,good luck

 

[-Job-]

I wouldn't say that there's exactly a "normal" number of hops in a trace route. Basically it depends on where you and the server you are trying to access are on the planet, as well as on your ISP and how the ISP is organized.
For me a tracert of www.google.com reported the following:

Tracing route to www.l.google.com [64.233.161.99]
over a maximum of 30 hops:

  1     2 ms     2 ms     3 ms  68-118-186-195.dhcp.nwtn.ct.charter.com [68.118.186.195]
  2    16 ms    10 ms     7 ms  10.24.0.1
  3    10 ms    11 ms    11 ms  24-151-8-33.static.nwtn.ct.charter.com [24.151.8.33]
  4    12 ms    10 ms    11 ms  sw65-1.dhcp.nwtn.ct.charter.com [172.20.98.65]
  5    13 ms    11 ms    14 ms  er-oc12.dhcp.nwtn.ct.charter.com [172.20.98.4]
  6    12 ms    13 ms    12 ms  ge-5-1-101.hsa1.Stamford1.Level3.net [63.208.120.1]
  7    15 ms    14 ms    13 ms  so-6-1-0.mp2.Stamford1.Level3.net [4.68.124.97]
  8    21 ms    18 ms    19 ms  as-3-0.bbr2.Washington1.Level3.net [4.68.128.206]
  9    18 ms    18 ms    20 ms  ae-22-52.car2.Washington1.Level3.net [4.68.121.51]
 10    21 ms    20 ms    20 ms  4.79.18.198
 11    19 ms    20 ms    21 ms  72.14.236.177
 12    21 ms    17 ms    17 ms  216.239.48.20
 13    20 ms    18 ms    21 ms  216.239.48.190
 14    21 ms    18 ms    23 ms  64.233.161.99

Trace complete.

Notice that the first hop, to my router (ip address 68-118-186-195) was the fastest one, as expected, since it's right there. BornLogician has a point, over heating is a common issue affecting routers as well, I've heard about it, as well as phones/microwave ovens, nearby routers, etc.

It's interesting that right now I'm studying routers and doing homework for my Networking class tomorrow, and so i just read about the MTU you mentioned, Maximum Transmission Unit, it's the maximum amount of data that a packet traveling in a given link can carry. If the MTU is bigger in the input links than it is in the output links, the router will have to fragment the packet, which means more processing, lower performance.

 

[-Job-]

Wait a second, i just noticed what you said about your first hop taking several seconds and timing out. That hop is to your router. Almost sounds like your router is getting more traffic than your DSL connection can support, and it's queueing up very quickly, dropping packets after that.
The reason why the router would do that might be problems with your DSL connection, or the router, or your DSL bandwidth isn't large enough to handle all the traffic, so the queues in your router grow larger and larger, until eventually they get full and all subsequent packets are dropped, causing your machine to resend some packets, which will be dropped.
Let me explain that more thoroughly. Most of the internet tasks you perform (Web, Mail, FTP) are over TCP which offers congestion control, so if packets are being dropped your machine doesn't keep resending them at the same rate, but slows down.
One scenario then is that the data coming into your router is more than the router can send out, meaning that packets are queued, and the queue is growing whenever you use your internet connection non-stop, for a while. It would then reach a point, maybe after the 30 minutes you mentioned, where the queue would be full, and any further packets are dropped. There's more than one versions of TCP which implement different congestion control algorithms like SS and AIMD (i have some homework problems on that), which regulate how much traffic is allowed to be sent after packets are dropped. Some of these set it to 1 packet, and other to more. In any case, it might take the 1 minute you mentioned for the congestion control to kick in and get traffic back to acceptable levels.
However once congestion goes away, TCP will increase the amount of traffic that can be sent, so the whole process would happen all over again, perodically. So traffic goes up until the router drops the packets, TCP detects this, reduces traffic, congestion goes away, router no longer drops packets, TCP increases allowed traffic, traffic goes up, rinse and repeat.
One important thing to note is that network congestion by itself won't cause you to lose connection, only slow it down, because TCP offers reliable data transfer, so lost packets are resent. The problem is that DNS uses UDP, which, unlike TCP, doesn't offer congestion control, nor reliable data transfer, so when the DNS packets are dropped, you'll get a "page not found error" for example. So it may not mean that you lost connection per se, just that you are getting a lot of network congestion.
Could be your bandwidth after all.

 

[DaveC426913]

just a couple of quick hardware ideas.. 2.4 giahertz phone in area cause same effect on my network.

Nope. That's why I tossed my 2.4GHz phone and put in a 5.4GHz phone.

When I did have the 2.4GHz I got a different effect - it would lose the wireless connection. It's not doing that.

I've also installed the router and dsl modem into a cooling system due to overheating new out of box. The cooling did help the dropoffs. Turning the router upsidedown to expose vents to prevent bottled up heat on chipsets. maybee. depending on the design. Simular to cooling cpu, overheat results in lockup? my router is a hotbox, which did need the cooling. As soon as I pushed the power button to the cooling system ms would drop to game server. simple ideas,good luck

It'd suck if I needed to set up some cooling system. I suppose I could start with some simple ways of allowing faster dissipation.

 

[DaveC426913]

Wait a second, i just noticed what you said about your first hop taking several seconds and timing out. That hop is to your router ...
... So it may not mean that you lost connection per se, just that you are getting a lot of network congestion.
Could be your bandwidth after all.

Yes. First hop is always "*...*...* Timed out." This is the case regardless of whether I'm having any problems.

Would that affect a tracert and a ping? When bandwidth stops, it stops completely.

Also, note that doing netspeed tests shows perfect bandwidth (200Mb down 500K up) at all times (except when bandwidth is down). I would think that lousy lines would casue general persistent degradation in bandwidth.

A "Repair" on my wireless connection seems to help, though it is not the magic bullet, as it often doesn't have any effect.

 

[russ_watters]

I have a very similar problem with Comcast at work. I really think it is an ISP issue, though. The thing is, it'll sometimes drop for about 10 seconds and then come back. But if it drops for longer, it doesn't come back on on its own and I need to reset the router.

 

[russ_watters]

Here's a normal tracert for me. I'll try to catch one when it drops.

Tracing route to www.l.google.com[/URL] [64.233.179.104]
over a maximum of 30 hops:

1 1 ms 1 ms 1 ms 192.168.2.1
2 * * * Request timed out.
3 11 ms 11 ms 11 ms ge-9-2-ur01.upperdublin.pa.panjde.comcast.net [6
8.86.217.117]
4 11 ms 11 ms 12 ms po10-ur01.plymouthmtng.pa.panjde.comcast.net [68
.86.209.166]
5 12 ms 11 ms 16 ms po20-ur02.norristown.pa.panjde.comcast.net [68.8
6.209.170]
6 14 ms 14 ms 14 ms po21-ar01.wallingford.pa.panjde.comcast.net [68.
86.209.106]
7 15 ms 15 ms 33 ms po90-ar01.401nbroadst.pa.panjde.comcast.net [68.
86.208.29]
8 19 ms 18 ms 16 ms 68.86.211.14
9 19 ms 18 ms 19 ms 12.119.53.49
10 28 ms 23 ms 40 ms tbr2-p012401.phlpa.ip.att.net [12.123.137.49]
11 25 ms 26 ms 38 ms tbr1-cl9.wswdc.ip.att.net [12.122.2.85]
12 22 ms 21 ms 21 ms 12.122.82.221
13 36 ms 37 ms 36 ms po3-1.ashcr2.Ashburn.opentransit.net [193.251.24
8.33]
14 36 ms 35 ms 36 ms so-1-0-0-0.atlcr1.Atlanta.opentransit.net [193.2
51.240.154]
15 37 ms 35 ms 37 ms google-us-peers-2.GW.opentransit.net [193.251.24
9.90]
16 36 ms 81 ms 226 ms 72.14.238.151
17 38 ms 37 ms 39 ms 72.14.238.178
18 36 ms 36 ms 38 ms 64.233.179.104

Trace complete.

 

[Anttech]

Anyway, they figured, to get around the problem, they lowered my MTU and did some other router setting stuff.

If your mtu is not the same as the ATM switch or whatever at your ISP's end you will get fragmentation! Check out what your isp recommends for this setting and stick to it

Yes. First hop is always "*...*...* Timed out." This is the case regardless of whether I'm having any problems.

means nothing, if you get a * it just means that the network device is not answering icmp echo requests (pings)

 

[Anttech]

Wait a second, i just noticed what you said about your first hop taking several seconds and timing out. That hop is to your router. Almost sounds like your router is getting more traffic than your DSL connection can support, and it's queueing up very quickly, dropping packets after that.
The reason why the router would do that might be problems with your DSL connection, or the router, or your DSL bandwidth isn't large enough to handle all the traffic, so the queues in your router grow larger and larger, until eventually they get full and all subsequent packets are dropped, causing your machine to resend some packets, which will be dropped.
Let me explain that more thoroughly. Most of the internet tasks you perform (Web, Mail, FTP) are over TCP which offers congestion control, so if packets are being dropped your machine doesn't keep resending them at the same rate, but slows down.
One scenario then is that the data coming into your router is more than the router can send out, meaning that packets are queued, and the queue is growing whenever you use your internet connection non-stop, for a while. It would then reach a point, maybe after the 30 minutes you mentioned, where the queue would be full, and any further packets are dropped. There's more than one versions of TCP which implement different congestion control algorithms like SS and AIMD (i have some homework problems on that), which regulate how much traffic is allowed to be sent after packets are dropped. Some of these set it to 1 packet, and other to more. In any case, it might take the 1 minute you mentioned for the congestion control to kick in and get traffic back to acceptable levels.
However once congestion goes away, TCP will increase the amount of traffic that can be sent, so the whole process would happen all over again, perodically. So traffic goes up until the router drops the packets, TCP detects this, reduces traffic, congestion goes away, router no longer drops packets, TCP increases allowed traffic, traffic goes up, rinse and repeat.
One important thing to note is that network congestion by itself won't cause you to lose connection, only slow it down, because TCP offers reliable data transfer, so lost packets are resent. The problem is that DNS uses UDP, which, unlike TCP, doesn't offer congestion control, nor reliable data transfer, so when the DNS packets are dropped, you'll get a "page not found error" for example. So it may not mean that you lost connection per se, just that you are getting a lot of network congestion.
Could be your bandwidth after all.

I don't think this is the problem, I can't see how a residential user could ever kill his FIFO ques at home... If it is network congestion it will be on the ISP network not at the router at home, unless that router is faulty. (or u have some sort of malware)
TCP does "Back off" persay, but it's not like single bus ethernet. It backs off because the SRTT time increases, so the RTO increases which is directly proportional to the time a source will wait to hear for a ACK... Anyway seems you are learing all about networking good for you :)

 

[DaveC426913]

Here's a normal tracert for me. I'll try to catch one when it drops.

Tracing route to www.l.google.com[/URL] [64.233.179.104]
over a maximum of 30 hops:

1 1 ms 1 ms 1 ms 192.168.2.1
2 * * * Request timed out.
3 11 ms 11 ms 11 ms ge-9-2-ur01.upperdublin.pa.panjde.comcast.net [6
8.86.217.117]
4 11 ms 11 ms 12 ms po10-ur01.plymouthmtng.pa.panjde.comcast.net [68
.86.209.166]
5 12 ms 11 ms 16 ms po20-ur02.norristown.pa.panjde.comcast.net [68.8
6.209.170]
6 14 ms 14 ms 14 ms po21-ar01.wallingford.pa.panjde.comcast.net [68.
86.209.106]
7 15 ms 15 ms 33 ms po90-ar01.401nbroadst.pa.panjde.comcast.net [68.
86.208.29]
8 19 ms 18 ms 16 ms 68.86.211.14
9 19 ms 18 ms 19 ms 12.119.53.49
10 28 ms 23 ms 40 ms tbr2-p012401.phlpa.ip.att.net [12.123.137.49]
11 25 ms 26 ms 38 ms tbr1-cl9.wswdc.ip.att.net [12.122.2.85]
12 22 ms 21 ms 21 ms 12.122.82.221
13 36 ms 37 ms 36 ms po3-1.ashcr2.Ashburn.opentransit.net [193.251.24
8.33]
14 36 ms 35 ms 36 ms so-1-0-0-0.atlcr1.Atlanta.opentransit.net [193.2
51.240.154]
15 37 ms 35 ms 37 ms google-us-peers-2.GW.opentransit.net [193.251.24
9.90]
16 36 ms 81 ms 226 ms 72.14.238.151
17 38 ms 37 ms 39 ms 72.14.238.178
18 36 ms 36 ms 38 ms 64.233.179.104

Trace complete.[/QUOTE]
This looks very much like mine, except that the last 8 or so hops are more like 100ms each.

 

[Anttech]

Do you use comcast?

why don't you post your traceroute?