Ensuring Fair-deal Service in Multi-core Processor Environment

Abstract

Present day multi-core processors integrate dozens of small processing cores with an on-chip network consisting of point-to-point links. The network enables communication between the processing cores and sharing of common resources such as memory and I/O controllers. In this work, we propose an arbitration scheme to ensure fair-deal service in accessing a chip’s shared resources by eliminating any bias in accessing a shared resource by a core based on its location in the multi-core chip. We propose using probabilistic arbitration combined with distance-based weights to achieve fair-deal service and overcome the limitation of conventional round-robin arbiter. We describe how nonlinear weights need to be used with probabilistic arbiters and propose three different arbitration weight metrics – constant weight, linearly increasing weight, and randomly increasing weight. By simply changing the arbitration of an on-chip router, we can create a simple, low-cost mechanism for achieving fair-deal service without requiring any additional buffers or virtual channels. We evaluate our arbitration scheme across a wide range of traffic patterns. In addition to providing fair-deal service, the proposed arbitration has additional benefits such as providing enhanced throughput, reduced latency, thus, providing a more stable network.