Feeling Your Website Through Different Backbones

A few months ago we moved our website from Alentus to Sungard (formly Inflow), with what we thought was great success. The latency between our office and our dedicated server was cut in half. When we accessed the server from various connections such as Qwest and Comcast our download rates were far improved. Then one day we received a call from a client saying that since we moved them into our new facility, they’ve received increased complaints of download failures.

Our initial response was one of skepticism because we were seeing improved performance, not a decrease. Using terminal services to connect to our other dedicated servers around the country I was able to reproduce the client’s experience. Apparently, about 70% of traffic around the country was having difficulty or decreased performance while accessing our servers. I then stumbled on a tool called Website Pulse.

Website Pulse monitors your website for performance and related issues from various locations internationally. In the chart below you can see the results of downloading a 22 MB file from our servers. The numbers on the left are seconds needed to download a file from Chicago over a backbone provided by nLayer, Level3, UUNET. The drop in seconds occurred when we moved our servers out of Sungard and into ViaWest in Denver, CO. The sudden increase on the 5th of October was an issue with the switch used in our rack.

Unknown to us, when Sungard acquired Inflow they also changed their policy in regards to capping bandwidth. Rather then allowing for bandwidth overages, and then surcharging the customer; Sungard’s policy was to cap bandwidth and not tell the customer of the performance hit.

Prior to this issue we didn’t realize various locations around the country experienced our websites in such a wide variety of speeds. With Website Pulse we receive alerts when any of our sites a falling below our acceptable load times.

It’s a great tool and I highly recommend using it for website monitoring.