Rather than try to estimate it, given we didn't know much about the software when we started, we just decided to do it first, then report on the cost afterwards. Doing the work in advance does increase the risk of not being funded for work done, but it has the advantage that we didn't have to speculate on the amount of time it took.

But it's true that we haven't done all the work yet. Of the remaining work, the most unpredictable is the amount of time it will take get the fileservers synching the way we will want, and if that will require more complicated software like CEPHS or glusterfs.