I've evaluated both through a proof-of-concept (current versions).
Both are good, both are equivalent in many respects.
DI is the easiest to use of all the major ETL tools. Easy to learn, easy to use, quick to build. Great user interface. Only vendor that can show side-by-side before and after of data going through transforms. Their data pipelining and parallelism support are good, but not as good as ASCL or INFA, so they won't scale up to the very high end as well. They include basic data profiling at no cost in all versions of the product. In all, I like the product very much.
DataStage has more pre-built features and much better enterprise-level manageability than DI. Easier to manage via an operations group, easier to manage multiple projects, easier to control developer and administrator access. It has poor debugging and data viewing features within the development environment. It is more expensive. If you buy the entire suite you get a real Metadata repository and Data Profiling and Data Quality, so they offer it all.
You must buy FirstLogic for profiling and quality, and another repository of choice for an enterprise MD repository if you use DI. But you will also pay a lot less if you do this best-of-breed approach with DI. The integration between the components in ASCL shows seams, which should not be the case in a single vendor product, but they acquired all through acquisitions.
Both offer real time. ASCL's is better if you are using the now-relabeled Mercator product. Otherwise the basic RT features are largely equivalent.