In recent years, the utilization of information systems has increased considerably in many different areas, and performance enhancement of these systems has become crucial.
In Real-Time systems in particular, where there are fundamental requirements such as low latency and temporal predictability, the management of shared resources like caches and memory bandwidth represents a real challenge. Inefficient management of these resources can cause interference between different applications, leading to degraded performance and reliability.
An important concept for real-time systems is Quality of Service (QoS), which guarantees certain levels of performance through a set of parameters and mechanisms. Furthermore, it ensures that specific requirements, such as latency, are met and guaranteed so that the system can be predictable and reliable.
To address these challenges and support QoS, ARM has created MPAM (Memory System Resource Partitioning and Monitoring). This technology allows us to manage shared resources by reducing their interference and contention. The main features of MPAM are the System Resourse Partitioning, which allows the various elements of the memory hierarchy to be partitioned, and the System Resourse Usage Monitoring, which allows us to monitor the usage of memory system resources.
This thesis will investigate the recent MPAM specification to create a self-Reconfiguring Architecture to support the Quality of Service of Real-Time systems.
This architecture, composed of modules that collaborate in an iterative pipeline mechanism, is named QoS Adaptive Reconfiguration Architecture (QARA).
QARA aims to analyze the performance of a partitioned system, leveraging a monitoring component, to be able to reconfigure the system parameters to improve performance for one or more partitions of the memory hierarchy.