Just a short post this time, but an interesting fact concerning different Internet Service Providers (ISPs) and their routing to/from other countries. I have a customer in Germany that has a remote office in France, connected via a site-to-site VPN. Around April last year the french office decided to change the ISP to a cheaper competitor that offers the same speed/bandwidth and therefore has no disadvantages… Well, I disagree.
The following image, take from MRTG with Routers2 and the mrtg-ping-probe tool, shows the latency (round-trip time, RTT) through the site-to-site VPN connection from the headquarter in Germany to the remote office in France. Obviously, in the mid of April the french office changed the ISP to another one which dramatically increased the RTT. After more than six months, they decided to go back to the old ISP which brought the expected low latency back.
This is basically an example that the “goodness” of an ISP does not only depend on its bandwidth offered to the customer, but also relies to the routing decisions and peerings to other ISPs, etc. Though the bandwidth was the same, the latency to/from Germany increased from about 400 ms peaks to 2000 ms peaks! You won’t ever work with that latency with Citrix or RDP!
Lessons learned: Prove the whole chain of processes before electing the cheapest ISP.