Do you have a question? Post it now! No Registration Necessary. Now with pictures!
January 20, 2006, 4:05 pm
rate this thread
I have a situation where a web site that is hosted by a ISP off-site seems
slow (this is our production environment). I have the same web site
on-site (use it for testing) which seems faster than the live site.
I have a developer that is not connected to our network or the ISP's
network. They have their own connection. Off hours when no one is using
the systems. He is saying that the test site is faster than the live site.
I'm trying to figure out where the problem is. Hosting ISP says it isn't
them, their systems are functioning normally. Our ISP connection out to the
internet is a fractional t-1 768k (different ISP than the ISP hosting the
web site) .
This ISP the one that has our t-1 says everything looks okay on their end
that we are not saturating our line. If you look at our Percent Utilization
using Solarwinds it is under 10% on both receive and transmit.
I guess my question is how do I see what is happening to traffic between me
and the ISP hosting our web site?
Our ISP that provides our T-1 mentioned something that doing a trace route
from us to our web site before it gets
to the company hosting the web site it goes through Williams Communication
(WCG) they say they know that there is a problem with latency there. I
believe WCG is the service provider for our ISP hosting the web site.
I know this is all a bit confusing but it is a difficult thing to
troubleshoot. I guess I don't know where to begin? The site does work just
seems to perform slow and our developer says that the pages aren't that big.
ANY HELP IS GREATLY APPRECIATED or pointing me in a direction to better
One would hope so - though this isn't a security problem. You have a
direct link to your local site. You are going to the "remote" site over
the Internet, which is not a dedicated or direct link for you.
That may be more relevant. Tell the developer to monitor the TCP Window
value and TCP Ack numbers (16th and 9th through 12th byte in the TCP
header) to detect connection stalls and packet drops. Received TCP sequence
numbers (5th to 8th byte in the TCP header) and IP packet lengths (3rd and
4th byte in the IP header) should step uniformly. Also look for indications
of IP fragmentation (7th and 8th byte in IP header - the F flag).
Not impossible. However, a traceroute (the original LBL version is using
UDP packets - the piece of sh1t that microsoft calls TRACERT uses ICMP
echos) is not representative of what is being used to carry the web page
(TCP). ICMP and UDP packets are often prioritized down as they are not
carrying "useful" data, and this may give false indications of transfer
Williams is a service provider. They have at least 7 /16s of IP space that
they lease out to others. Searching for WCG, or 'Williams Communication'
in the 'news.admin.net-abuse.* newsgroups will provide more details.
"pages aren't that big" doesn't say anything useful. What is on the pages,
and where is it being sourced from may be a lot more significant. Trying
to troubleshoot it is not easy, as you are depending on multiple hops
between your web page and those who wish to view it. You really don't
have control on these intermediate hops. About all you can do it test
the connection speed from various sites and graph the results. If you
detect a problem (for example, excessive delay when using WCG IP space)
your only solution may be to use a different provider that doesn't
route over the problem provider.