While MPLS typically provides 99% uptime, multiplied across 10 branches, even short outages add up, says Bob Fourness, computer operations manager at Hydrite. He wrestles with two scenarios: sporadic outages, where a branch goes down two to three times per year; and chronic problems, where there is an outage every week for two to three months.

While the company has added DSL backup on some of its remote routers, the failover processes takes two to five minutes, disconnecting the company's IBM 5250 terminal sessions, meaning users have to log back in when the connection is re-established and, oftentimes, have to call IT.

"We have a number of plants that work three shifts, so if that happened at night it would involve a call to one our people outside of normal hours to reset printers and that sort of thing," Fourness says.

Besides the terminal traffic to the company's CRM app on an IBM iSeries AS400, the network supports PC backups, Internet access (backhauled to HQ), Lotus Notes replication, terminal emulation and file sharing.

So reliability was a driver, but so was the need for additional bandwidth, in particular at those two main branches.