SDKV: A Smart and Distributed Key-Value Store for the Edge-Cloud Continuum
Juan Aznar-Poveda, Tobias Pockstaller, Thomas Fahringer, Stefan Pedratscher, Zahra Najafabadi Samani · 2023
Many time-critical and data-intensive distributed applications for the computing continuum depend on low-latency, scalable, and highly available distributed key value storages. In this paper, we introduce SDKV, a scalable -Smart and Distributed Key-Value- store for the Edge-Cloud continuum to automatically place data in close proximity to clients resulting in low response times. The clients of SDKV can influence data availability and access latency by specifying the number of replicas and the desired level of data consistency (strong or eventual) on a per key-value pair basis, which favors the support of a wide range of applications. Results reveal that for different workloads and client access behaviors, SDKV outperforms existing distributed data storages and their data placement algorithms by 12--69% for both consistency models. Moreover, the proposed placement algorithm of SDKV provides fast decision times and scales linearly with the number of keys.