KVCG: a heterogeneous key-value store for skewed workloads
KVCG: a heterogeneous key-value store for skewed workloads
复制标题
DOI：
10.1145/3456727.3463779
复制
发表时间：
2021-06
期刊：
影响因子：
--
通讯作者：
dePaul Miller;J. Nelson;Ahmed Hassan;R. Palmieri
中科院分区：
文献类型：
--
作者：
dePaul Miller;J. Nelson;Ahmed Hassan;R. Palmieri
我们提出了 KVCG，一种新颖的异构键值存储，其主要目标是以亚毫秒延迟为针对频繁访问（热）键的客户端请求提供服务，并以高吞吐量服务针对不太频繁访问（冷）键的请求。为了实现这一目标，KVCG 部署了一种架构，其中热键请求被路由到由 CPU 线程操作的软件缓存，而其余部分则被卸载到针对现代 GPU 设备上执行而优化的数据存储库。冷/热分区是在运行时通过使用传入工作负载进行训练的模型完成的。与最先进的竞争对手相比，我们的延迟提高了 34 倍。
We present KVCG, a novel heterogeneous key-value store whose primary objective is to serve client requests targeting frequently accessed (hot) keys at sub-millisecond latency and requests targeting less frequently accessed (cold) keys with high throughput. To accomplish this goal, KVCG deploys an architecture where requests on hot keys are routed to a software cache operated by CPU threads, while the remainder are offloaded to a data repository optimized for execution on modern GPU devices. Cold/hot partitioning is done at runtime through a model trained with the incoming workload. Against a state-of-the-art competitor, we obtain up to 34x improvement in latency.