Abstract:
Edge computing is a new computing paradigm that has emerged to offload computation and storage to edge devices in order to get a shorter response time and more efficient processing. Nevertheless, adopting a consistency scheme that can conserve multiple replicas while guaranteeing a good level of consistency is an open issue. Moreover, a data store with only one consistency level is not suitable for applications that have different consistency requirements. In this paper, we propose MinidoteACE, a new adaptive consistency system which is an amelioration version of Minidote a causally consistent system for edge applications. Unlike Minidote which supports only causal consistency, our model allows applications to run also queries with stronger consistency guarantees. Experimental evaluations show that throughput decreases only by 3.5% to 10% when replacing causal operation with a strong operation. However, update latency increase significantly for strong operations up to three times for 25% update workload.