First off, I'm guessing that you might need something more complex than a simple LP solver. Most cell phone services have breakpoints where you may want to switch from one service to another based on call length, frequency, time of day, etc. This switching implies the need for integer variables which means you might need a MILP (mixed integer linear programming) solver. (If all of your cost functions and constraints are convex, you could get by with an LP solver, but that's getting ahead of ourselves a little). The good news is that there are also open source and affordable MILP solvers out there as well.
I'd start with LP SOLVE or SYMPHONY. Check out the COIN-OR site here for some useful background info.
In response to your enhanced problem description, I think you could simply take each of the 50 users and cost out each plan then apply each of the options individually. With n users and m possible plans and p possible options, you need to look at m*p options for each user - but that's kind of boring.
A more interesting question from the user perspective is: Where are the break points between plans? Can you define indifference curves - combinations of usage where the user would be indifferent between two plans? This question could be address mathematically probably using some linear algebra techniques, but there's really not an objective function so it doesn't seem like a MILP.
Another interesting question from the provider perspective - how to set plans to maximize profits? Here you could apply some optimization if you take your 50 users to be representative of the population. You would need to put a cap on the total cost for a user and add costs to get profits, but I think a formulation is possible.