Parallel computing research in the area of nonlinear optimization has been extremely intense during the last decade. One of the perspectives for this research is to parallelize the linear algebra operations that arise in numerical optimization, and thus to consider the impact that the most recent advances in parallel linear algebra can have in building efficient high-performance software for nonlinear optimization. Bringing the high-quality parallel linear algebra algorithms and software, produced in the last decade, into the field of nonlinear optimization is really a crucial issue to deal with. This is because the linear algebra needs in optimization strongly depend on the kind of problem and/or method one considers, and therefore, for each algorithm, it is necessary to thoroughly analyze the parallel numerical linear algebra strategy to be adopted. By considering a few general classes of methods, well representative for the different aspects of the overall problem, some of the crucial linear algebra kernels, some related existing parallel software and/or algorithms, as well as some specific linear algebra parallelization issues, unresolved or little considered, will be pointed out. Finally, some issues related to parallel preconditioning will also be addressed.