The pessimism of this computation comes directly from the fact that it considers a constant cost for every possible preemption, and this constant cost is assumed to be the maximum possible cost. That is, this approach is not sensitive to the preemption cost pattern of the task. As it was claimed in the abstract, using this additional information (the tasks preemption cost pattern) enables us to derive a much more accurate upper-bound. This second technique is described in Algorithm 1, and a detailed explanation is provided below.