I have run both DSL and Cable over the last year on a NT4 trading platform. I considered a number of configurations and ended up using two Asante routers with different IP (gateway) addresses hooked to my office network. In my workstation, I use one NIC assigned two IP addresses and two gateway addresses. Technically, NT will automatically switch from one gateway (DSL) to the other (Cable) when one service fails. The problem is that failure here means the unlikely failure of the local network or router itself and not the internet path. When I want to switch between DSL and Cable, I use two small batch programs on my desktop to flip the primary gateway.
route delete 0.0.0.0
route add 0.0.0.0 mask 0.0.0.0 xxx.xxx.xxx.xxx
Where xxx.xxx.xxx.xxx is the IP of the gateway (router). This manual configuration gives me control over the timing of the switch over.
While the Nexland router looks like an affordable backup product , it doesn't fill my needs as a trader. I typically have a data feed and 3 to 5 broker applications running. All of these are connected to different IPs with different routes. A failure is any unacceptable condition on any route lasting more than 10 to 20 seconds. Unacceptable performance includes high latency (ping > 500ms), lost packets (> 10%) or failure of any link in the route to my supplier's server. As far as I know, the Nexland device will look for hard failure of the path to one server every 20 seconds. I doubt that it would efficiently evaluate multiple or even degraded paths.
I could write a piece of software to monitor each path; however, the switch over itself has some undesirable results. When a connection is lost, my data feed and broker interfaces go into a time-out mode (not visible) and then each application attempts to relog-in. Unfortunately, during the time-out period, you may not even know you are down. The total process takes up to 90 seconds before everything is back on-line. If the route failure is common to both ISP paths (for example, the broker's server), then an automated system would alternatively switch between ISPs trying to find a connection. Thus, the loss of one of five applications takes my whole platform down. For these reasons, I prefer to manually switch between ISPs when I have evaluated the nature of the problem using a route tool like ping plotter.