How to run a "traceroute"
No problems listed on our network status
page and your access is still slower than normal? There is a chance that you may
experience slow connectivity if the gateway between your backbone and our backbone does
not have sufficient capacity. That point where the data changes backbones is referred to
as a "peering point". Traditionally, peering points are the number one
source of slowness on the Internet. This is because peering arrangements are usually
done such that neither company pays the other. As a result, if that peering
connection gets slow due to insufficient capacity or low quality hardware, neither
backbone company thinks they should be the one to fix it. Backbone providers commonly
used by DSL providers and cable network providers often have those problems due to over
selling their available bandwidth.
If you're experiencing slow access to your server, there is a simple
way to check if it is a peering point or something else.
Open a MS-DOS prompt and run the following command:
Microsoft(R) Windows 98
(C)Copyright Microsoft Corp 1981-1998.
This will trace the route that your data takes to our datacenter and tell you where the
slowness begins. Each line of the output represents another "hop" that the data
has to go through to reach its destination. A sample output would look like the
Tracing route to schmolie.com [188.8.131.52]
over a maximum of 30 hops:
1 2 ms 1 ms 1 ms 184.108.40.206
2 95 ms 97 ms 94 ms atm11-0-0106.cr-1.usstls.savvis.net [220.127.116.11]
3 95 ms 94 ms 95 ms 500.POS5-0-0.GW2.STL1.alter.net [18.104.22.168]
4 94 ms 95 ms 94 ms 102.ATM2-0.XR2.STL1.ALTER.NET [22.214.171.124]
5 95 ms 95 ms 95 ms 192.at-1-1-0.TR2.STL3.ALTER.NET [126.96.36.199]
6 128 ms 128 ms 127 ms 134.at-6-1-0.TR2.NYC9.ALTER.NET [188.8.131.52]
7 140 ms 128 ms 134 ms 0.so-3-0-0.XR2.NYC9.ALTER.NET [184.108.40.206]
8 128 ms 128 ms 128 ms 180.ATM7-0.BR1.NYC9.ALTER.NET [220.127.116.11]
9 129 ms 128 ms 130 ms 18.104.22.168
10 155 ms 154 ms 144 ms serial.cambridge.ma.psi.net [22.214.171.124]
11 147 ms 155 ms 156 ms rc6.se.us.psi.net [126.96.36.199]
12 162 ms 156 ms 146 ms 188.8.131.52
13 149 ms 146 ms 146 ms 184.108.40.206
14 158 ms 168 ms 159 ms schmolie.com [220.127.116.11]
Note: Our network has been upgraded considerably since this
traceroute was done, but it still makes a great example. We now have an average latency of
20ms between our datacenter and the Internet, much better than the 95ms shown above.
Now, in this example trace route from our Walla Walla datacenter to
our former PSINet datacenter, there are no congested points. If there were, you would see
the "*" symbol instead of the numbers on the left or the numbers would be much
higher. You can tell where the trouble begins by looking at the names next to the the
"*" symbols or the high numbers. The first line that has high numbers or
"*" symbols is where the trouble begins. If the "hop" before or
after the problem hop is a different backbone, then most likely a peering point is having
problems. You can tell which backbone it is by looking at the domain name next to
the IP address. If it starts having problems right at the first few "hops",
then it is most likely a problem with your ISP or your ISP's backbone company. In
some cases, intermittent high numbers may manifest themselves at later hops due to one of
the first few hops only having intermittent problems. The only way to tell for sure
where the problem starts is to run a series of traceroutes over the same route one after
another until you are certain which hop is causing the intermittent problems. The
reason for this is rather simple: to get to point "D", you have to go through
points "A", "B", and "C". If point "A",
"B", or "C" is having intermittent problems, those same problems will
be seen for all hops after the problem hop, though intermittently.
You may be wondering what would be considered a "high"
number when doing a traceroute. That primarily depends on what you are using to
connect to the Internet. For instance, your average modem will add 120ms to 300ms
depending on the phone line conditions and the make/model of the your modem and your ISPs
modem. Most broadband connections would add between 10ms and 100ms depending on the
technology in use. The exception is sattelite broadband Internet connections: they
typically see between 400ms and 2000ms depending on the hardware used and the transmission
distance. Whatever your connection technology, any latency associated with your
chosen technology will show itself on the first hop. In most cases, that first hop
indicates the latency of the connection between you and your ISP. Any numbers that are
more than 100ms to 200ms more than your chosen technology's latency could be considered
"high" by most standards.
The solution to a problem with your ISP is to use a different
ISP. The solution to the peering point problem is for one backbone company to pay the
other as if it were a regular customer. That causes a contractual obligation for the paid
backbone company to fix any problems and add capacity if needed. We choose our
backbone connections carefully with regard to how they handle peering point issues to
ensure our customers get good connections. Another possible solution to a peering
point problem is to email your traceroute to email@example.com
and detail the trouble you're having. Depending on the problem, we may be able to route
around the slow portion of the Internet.
Please note that some "Universal Pipe" accounts have a
data rate determined by what you choose to pay for. If you have a rate limited
Universal Pipe account and experience high numbers at the last hop of the traceroute, you
should probably upgrade your Universal Pipe to a higher
data rate or reduce your bandwidth usage.