We describe a collection of linear systems generated during the iterations of an interior-point method for convex quadratic optimization. As the iteration index grows, the systems may become increasingly ill conditioned. Each system represents the linearization of Newton's equations about an iterate and for a certain value of the barrier parameter. A salient feature of the collection is that each system comes in the form of its block components. Thus, it is possible to use the collection to benchmark direct or iterative solvers on various formulations of the linearized Newton equations. Matlab tools are supplied to facilitate forming those formulations.