Real-Time Scheduling of Hybrid Systems using Credit-Controlled Static-Priority Arbitration

The convergence of application domains in new systems-on-chip results in hybrid systems with many intellectual property components with a mix of soft and hard real-time requirements. Resources, such as memories and interconnect , are shared between requestors to reduce cost. However, resource sharing introduces interference between the sharing components, referred to as requestors, making it difficult to satisfy their real-time requirements. Existing arbiters do not efficiently satisfy the requirements of the requestors in these systems, as they couple rate or allocation granularity to latency, or cannot run at sufficiently high speeds in hardware with a low-cost implementation. The contribution of this document is an arbiter called Credit-Controlled Static-Priority (CCSP) consisting of a rate regulator and a static-priority scheduler. The rate regulator isolates requestors by regulating the amount of provided service. Regulating provided service, as opposed to regulating requested service has two benefits: 1) the implementation of the regulator is less complex, and 2) the amount of work associated with a particular request does not have to be known. We show that CCSP belongs to the class of latency-rate servers and guarantees the allocated service rate within a maximum latency, required by hard real-time requestors. We present an implementation of the arbiter in the context of a DDR2 SDRAM controller that has been efficiently integrated into the network interface of a network-on-chip. The implementation supports service allocation with negligible over-allocation, and its area is small and scales linearly with the number of re-questors. An instance with six ports runs at 250 MHz and requires 0.0175 mm 2 in CMOS090LP. Unclassified Conclusions: In this document, we present an arbiter called Credit-Controlled Static-Priority (CCSP) for scheduling access to resources, such as interconnect and memories in systems-on-chip. CCSP resembles an arbiter with a rate regulator that enforces a (σ, ρ) constraint on requested service together with a static-priority scheduler. However, instead of enforcing an upper bound on requested service, CCSP enforces an upper bound on provided service. Regulating provided service, as opposed to regulating requested service has two benefits: 1) the implementation of the regulator is less complex, and 2) the amount of work associated with a particular request does not have to be known. We show that CCSP enjoys these benefits, without increasing worst-case latency or buffering, compared to an arbiter regulating requested service. We furthermore show that CCSP belongs to the class of latency-rate (LR) servers and guarantees the allocated service rate …