How many problems have been caused by not having a clustered solution previously?
My suggestion is to weigh the cost of setting up the cluster (in terms of time, additional license fees, increased administration resources, etc.) versus the cost of a failure.
If a failure is merely a nuisance that causes someone a bit of extra work on rare occassions then I would say you have no need for clustering, other than to be

.
On the other hand, if a failure is very costly to the business then I would say the time and money would be well spent investing in clustering.