The question of which and how a particular class of entangled resource states (known as graph states) can be used for measurement based quantum computation (MBQC) recently gave rise to the notion of Flow and its generalisation gFlow. That is a causal structure for measurements guaranteeing deterministic computation. Furthermore, gFlow has proven itself to be a powerful tool in studying the difference between the measurement-based and circuit models for quantum computing, as well as analysing cryptographic protocols. On the other hand, entanglement is known to play a crucial role in MBQC. In this paper we first show how gFlow can be used to directly give a bound on the classical simulation of an MBQC. Our method offers an interpretation of the gFlow as showing how information flows through a computation, giving rise to an information light cone. We then establish a link between entanglement and the existence of gFlow for a graph state. We show that the gFlow can be used to bound the entanglement width and what we call the structural entanglement of a graph state. In turn this gives another method relating the gFlow to bounds on how efficiently a computation can be simulated classically. These two methods of getting bounds on the difficulty of classical simulation are different and complementary and several known results follow. In particular known relations between the MBQC and the circuit model allow these results to be translated across models.Measurement Based Quantum Computing (MBQC) [1] has attracted attention recently for its potential towards the realisation of a quantum computer, its role in understanding the power and significance of entanglement for computation [2,3], and that it plays a key role in the development of cryptographic protocols [4,5]. In MBQC one starts off with a large multiparty entangled resource state and the computation is driven by a series of local measurements, the choice of which can depend on the result of previous measurements in the series. The formal language for MBQC was jointly developed by Prakash Panangaden in [6]. In this work we are interested in the question of how to recognise or characterise a 'good' resource for measurement based quantum computing. Given the fact that after the generation of the state, all operations are local, it