https://i.imgur.com/d1XnZ0i.png
We have a 400Mbps commit and 1Gbps burst from Century Link in Seattle. We also have access to a 1Gbps “business” line from Comcast at the same building.
We noticed a problem when systems on the CL circuit were taking a LOT longer to run updates and pull data from online compared to the Comcast circuit.
The above link is a table of test http results from various data centers in seattle and across the US.
The CL circuit is basically garbage to every single site.
I setup a iperf3 server on vultr in seattle and on the Comcast circuit I got 800Mbps. On the CL circuit I got ~180-200Mbps. I captured wireshark of the transfer. The Comcast connection looked stable. The CL circuit looked like a jagged sawtooth with >13k dupe acks.
We escalated to CL. Their response “we ran a L2 test and got 488Mbps on the circuit, the problem is on your end”.
1) We plugged directly into our Cisco 1002-X router with it straight to the CL circuit and nothing else. Local test to another client on router is fine. 2) how do you conclude that a 488Mbps circuit test result is “good” on a 1Gbps burstable line?
Like wtf? I don’t typically work with ISPs much. I just help run the network for our labs and interface with our enterprise networking people which are mediating and investigating with CL.
But holy f%*%. Why are ISPs alway such a PITA when trying to help diagnose issues when you present them with all kinds of data that there is a problem? They just balk at it, look to see if it auto negotiated to 1Gbps or something stupid and say “line is good”. Like what in the hell.
Our enterprise networking is escalating the issue with CL right now. So hopefully we get someone that isn’t a compete derp. Maybe I’m completely wrong...but just annoyed frustrated and mainly wanted to rant.
No comments:
Post a Comment