We considered something like the Lambda Architecture and rejected it. We aren’t a large enough team to adequately support both a map/reduce and streaming framework. So far, we haven’t felt any ill effects Storm has worked out quite nicely playing both roles.
However, there are certain problems that Storm isn’t as well suited to that a system like Hadoop is that we have yet to need to do. We have one of those coming up and we’re interested to see where we end up.