The recent OWL 2 W3C recommendation includes the lightweight ontology language OWL EL which is semantically based on an extension of the EL++ description logic (DL). It is widely assumed that inferencing in OWL EL is possible in polynomial time, but it is not obvious how to extend existing reasoning procedures for EL++ accordingly. We set out to close this gap by developing inferencing methods for SROEL(⊓, ×) – a DL that subsumes the main features of OWL EL. We present a framework for studying materialisation calculi based on datalog, and we use it to investigate the resource requirements for inferencing. We can show that certain SROEL(⊓, ×) feature combinations must lead to increased space upper bounds in any materialisation calculus, suggesting that efficient implementations are easier to obtain for suitably chosen fragments of SROEL(⊓, ×).