OSDI文章分享，论文（Fast RDMA-based Ordered Key-Value Store using Remote Learned Cache）的结构化文摘

最新推荐文章于 2024-10-28 16:00:09 发布

原创最新推荐文章于 2024-10-28 16:00:09 发布 · 496 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#操作系统 #rdma #机器学习 #缓存

一分钟看懂操作系统顶会论文OSDI 专栏收录该内容

34 篇文章 ¥9.90 ¥99.00

订阅专栏

超级会员免费看

XSTORE是一个利用RDMA的高效有序键值存储系统，采用远程学习缓存来解决树形索引遍历慢的问题。通过将机器学习模型作为理想的缓存结构，它在服务器端保留树形索引处理动态工作负载，客户端的学会缓存处理静态工作负载。这种设计允许过时的学习缓存仍能预测查找键的位置，并通过验证机制和推测执行确保正确性。实验表明，XSTORE在只读请求性能上超越了现有RDMA基键值存储，对于插入工作负载也有显著的吞吐量提升。

Authors:

Xingda Wei, Rong Chen, and Haibo Chen, Shanghai Jiao Tong University

Abstract:

RDMA (Remote Direct Memory Access) has gained considerable interests in network-attached in-memory key-value stores. However, traversing the remote tree-based index in ordered stores with RDMA becomes a critical obstacle, causing an order-of-magnitude slowdown and limited scalability due to multiple roundtrips. Using index cache with conventional wisdom—caching partial data and traversing them locally—usually leads to limited effect because of unavoidable capacity misses, massive random accesses, and costly cache invalidations.

We argue that the machine learning (ML) model is a perfect cache str

了解本专栏