As some others suggested I would also recommend to try to make an experiment. Identify one part of your software suite you think you can build, operate and do the maintenance for. Aim to do it as fast as possible. Test if you can resolve the issues, you have with your vendor, if you do it yourself. If that fails, it propbably isnt a good idea to go full scale on it. Again: aim to get some real world experience with your own build software AS FAST (and cheap) as possible. Dont go for a perfect solution.