I. INTRODUCTION Partitioned architectures (PAs) allow the safe integration of applications of different criticality levels on the same platform, reducing the development, verification and integration costs. PAs rely on partitioning mechanisms at the platform level to ensure temporal and spatial separation between applications of different criticality levels. With PAs, each application is running in its own partition. Spatial partitioning protects the private data or devices of an application in a partition from being tampered with, by another application. Temporal partitioning ensures that an applications access to shared resources is not affected by applications in other partitions.PAs have been successfully used in several industries, including automotive and avionics. For example, in the avionics area, platform level separation mechanisms are described in the ARINC 653 software specification, also called Integrated Modular Avionics [3]. Recently, the European Space Agency (ESA) and the National Aeronautics and Space Administration (NASA) have also shown interest in PAs, as a way to "manage the growth of mission function implemented in the on-board software" [8], and as intermediate step to introducing multicore processors in spacecraft computers [7].In [6], we have addressed the optimization of PAs for hard real-time applications, focusing on finding schedulable implementations that minimize the development and certification costs. In this paper we are not interested in the issue of cost minimization, but in supporting soft real-time applications that share the same PA with critical hard real-time applications. The advantage of a PA is that it allows the integration of mixed-criticality applications, including non-critical and soft real-time applications, onto the same platform. Our proposed optimization approach determines an implementation such that all hard real-time applications are schedulable and the quality of service of the soft real-time tasks is maximized.