[Submitted on 28 Jul 2026]

View PDF HTML (experimental)

Abstract:RDMA 已在解聚記憶體資料庫中實現高速資料存取與低延遲通訊。雖然已提出各種最佳化技術，以 RDMA 加速此情境下的交易，但雙邊 RDMA 因涉及遠端 CPU，長期以來被低估，研究多偏向單邊 RDMA。然而，單邊 RDMA 的大量使用帶來根本性的限制。其有限的 API 無法表達複雜的系統功能，例如防飢餓、基於優先權的排程與搶占，而這些功能在併行控制協定中至關重要。此外，索引需要多次網路往返，造成網路放大。
In this work, we revisit the long-standing debate between one-sided RDMA and two-sided RDMA in the context of disaggregated memory databases. We present Lotus, which addresses the conventional limitation of two-sided RDMA, i.e., CPU bottlenecks in memory servers, by leveraging the rich functionality of two-sided RDMA with two key optimization techniques: (1) lightweight caching and (2) efficient batching. Lotus demonstrates that limited CPU resources in memory servers, when intelligently utilized, can transform a perceived weakness into a significant advantage. Our experimental study shows that Lotus achieves up to 8.2$\times$ higher throughput and 42.9$\times$ lower p999 tail latency than state-of-the-art one-sided RDMA-based approaches in YCSB benchmark.

Submission history

From: Hokeun Cha [view email]
[v1] Tue, 28 Jul 2026 20:04:01 UTC (1,012 KB)