Resource Arbitration

Abstract

The previous chapter presented a memory controller back-end that makes an SDRAM into a predictable resource, corresponding to the first part of our approach to predictability. The second part of the approach, which is the topic of this chapter, considers sharing this resource among multiple requestors in a predictable manner. The context of this problem was previously shown in Fig. 1.7, where requests arrive in a Request Buffer in front of a resource arbiter and responses are returned in a Response Buffer. Resource arbitration with real-time requirements is in no way a new research field. In fact, research has been conducted in this field during more than half a century already and there exists a plethora of different arbiters. Still, new applications and emerging technologies like heterogeneous multi-core System-on-Chips (SoCs) continue to change the requirements, as they need small and fast arbiters that cater to diverse requirements without wasting scarce resource capacity