In this presentation we will see a general background over iterative linear solvers and the bottlenecks prohibiting reducing the approximately 80 % of simulation time. This time is passed only to solve linear systems. These methods are widely used in both scientific and industrial fields. Parallelising these methods such that they are applied on supercomputers is natural since they rely on a matrix vector product and vector dot product. However, these operations are constrained by the communication time between processors. We will see a brief introduction to HPC. And finally, we will see how we could reduce the global communication by proposing multiple search directions variants.