Summary: Recent development in Extended Finite Element Method (XFEM) opened new avenues thorough crack propagation problems. However, it ability to predict crack path in micro scale medium of a real porous rock is questionable. In this work, we compare two strategies and introduce one as the best strategy to use the XFEM for such a purpose in ABAQUS Software. We demonstrate our claim by comparing numerical results with analytical solution and experimental test. Introduction: Crack growth has always been one of the major challenges in the rock mechanics. Although pores, joints and fractures are the most critical structures controlling cracks initiation and propagation, their spatial distribution and geometrical effects are not still well-understood. In this study, we aim to numerically model the crack growth in a real porous rock. Methodology and Approaches: We use the extended finite element method (XFEM), which has recently attracted more attentions due to its ability to estimate the discontinuous deformation field by using special shape functions. Since direct use of the XFEM does not lead to an accurate result, two different strategies are considered for applying the XFEM on a porous model to simulate the crack propagation. Results and Conclusions: Our results showed that applying several different partitions and enrich them individually lead more logical results than to allocate reduced elastic modulus to porosity. We used this strategy to evaluate the XFEM both analytically and experimentally as a possible numerical solution. Thus, two simple models were constructed, both numerically and experimentally (Granite): i. a sample with one void and one crack, and ii. a sample with two voids and one crack between them. Analytical solutions for the stress intensity factor revealed that the XFEM modeling can compute this parameter with an error less than 5%. On the other hand, experimental results showed that the XFEM with partitioning strategy can predict the correct crack growth path comparative to the experimental results. Accordingly, digital images of Berea sandstone were used as a real reservoir rock and, then, this method was implemented to simulate multi-crack propagation through the exact medium of rock.