Rapid increase of storage and network bandwidth incurs higher CPU consumption in modern data systems. This phenomenon is particularly evident for log-structured merged key-value stores (LSM-KVS), which rely on resource-intensive background operations to flush and compact disk data. While extensive research has been conducted to reduce the CPU overhead of background compaction, less attention has been paid to background flushing, which can also consume a significant amount of valuable CPU cycles and disrupt CPU caches, ultimately impacting overall performance. In this paper, we propose DFlush, a novel solution that uses DPUs to offload background flush operations to reduce its CPU cost. DPUs are an appealing choice for this goal due to their cost-effectiveness, ease of programming, and widespread deployment. However, their complex hardware architecture requires careful design of both the data and control planes. To fully harness the DPU's capabilities, DFlush decomposes a flush job into fine-grained steps, mapped them to DPU hardware units, and accelerates them through pipeline, data, and channel parallelism, ensuring data-plane efficiency. It also introduces an adaptive control plane that dynamically schedules flush jobs from different LSM-KVS instances based on their priority, reducing write stall and tail latency. Our experiments on a real DPU platform with an industrial-grade LSM-KVS show that DFlush delivers higher throughput, significantly lower tail latency, and saves up to dozens of CPU cores per LSM-KVS server while reducing energy consumption.

LSM-based key-value stores suffer from sub-optimal performance due to their slow and heavy background compactions. The compaction brings severe CPU and network overhead on high-speed disaggregated storage. This article further reveals that data-intensive compression in compaction consumes a significant portion of CPU power. Moreover, the multi-threaded compactions cause substantial CPU contention and network traffic during high-load periods. Based on the above observations, we propose fine-grained dynamical compaction offloading by leveraging the modern Data Processing Unit (DPU) to alleviate the CPU and network overhead. To achieve this, we first customized a file system to enable efficient data access for DPU. We then leverage the Arm cores on the DPU to meet the burst CPU and network requirements to reduce resource contention and data movement. We further employ dedicated hardware-based accelerators on the DPU to speed up the compression in compactions. We integrate our DPU-offloaded compaction with RocksDB and evaluate it with NVIDIA’s latest Bluefield-2 DPU on a real system. The evaluation shows that the DPU is an effective solution to solve the CPU bottleneck and reduce data traffic of compaction. The results show that compaction performance is accelerated by 2.86 to 4.03 times, system write and read throughput is improved by up to 3.2 times and 1.4 times respectively, and host CPU contention and network traffic are effectively reduced compared to the fine-tuned CPU-only baseline.

Learned Index, which utilizes effective machine learning models to accelerate locating sorted data positions, has gained increasing attention in many big data scenarios. Using efficient learned models, the learned indexes build large nodes and flat structures, thereby greatly improving the performance. However, most of the state-of-the-art learned indexes are designed for DRAM, and there is hence an urgent need to enable high-performance learned indexes for emerging Non-Volatile Memory (NVM). In this article, we first evaluate and analyze the performance of the existing learned indexes on NVM. We discover that these learned indexes encounter severe write amplification and write performance degradation due to the requirements of maintaining large sorted/semi-sorted data nodes. To tackle the problems, we propose a novel three-tiered architecture of write-optimized persistent learned index, which is named WIPE , by adopting unsorted fine-granularity data nodes to achieve high write performance on NVM. Thereinto, we devise a new root node construction algorithm to accelerate searching numerous small data nodes. The algorithm ensures stable flat structure and high read performance in large-size datasets by introducing an intermediate layer (i.e., index nodes) and achieving accurate prediction of index node positions from the root node. Our extensive experiments on Intel DCPMM show that WIPE can improve write throughput and read throughput by up to 3.9× and 7×, respectively, compared to the state-of-the-art learned indexes. Also, WIPE can recover from a system crash in ∼ 18 ms. WIPE is free as an open-source software package. 1

LSM-based Key-value stores suffer from sub-optimal performance due to their slow and heavy background compactions. The compaction overhead shifts to the CPU as the storage performance continuously increases. This paper further reveals that data-intensive compression in compaction consumes a significant portion of CPU power. Moreover, the multi-threaded compactions cause substantial CPU contention during high-load periods. Based on the above observations, we propose fine-grained dynamical compaction offloading by leveraging the modern Data Processing Unit (DPU) to alleviate the CPU overhead. To achieve this, we first employ dedicated hardware-based accelerators on the DPU to speed up the compression in compactions. We then leverage the Arm cores on the DPU to meet the burst CPU requirements to reduce resource contention. We integrate our DPU-offloaded compaction with RocksDB and evaluate it with NVIDIA’s latest Bluefield-2 DPU on a real system. The evaluation shows that the DPU is an effective solution to solve the CPU bottleneck of compaction. The results show that compaction performance is accelerated by 2.86 to 4.03 times, system write and read throughput is improved by up to 3.2 times and 1.4 times respectively, and host CPU contention is effectively reduced compared to the fine-tuned CPU-only baseline.

Popular LSM-tree based key-value stores suffer from suboptimal and unpredictable performance due to write amplification and write stalls that cause application performance to periodically drop to nearly zero. Our preliminary experimental studies reveal that (1) write stalls mainly stem from the significantly large amount of data involved in each compaction between$L_{0}$-$L_{1}$(i.e., the first two levels of LSM-tree), and (2) write amplification increases with the depth of LSM-trees. Existing work mainly focus on reducing write amplification, while only a couple of them target mitigating write stalls. In this paper, we exploit unique features of non-volatile memory (NVM) to address these two limitations and propose TriangleKV, a new LSM-tree based persistent KV store with multi-tier DRAM-NVM-SSD storage. TriangleKV's design principles include performing smaller and cheaper$L_{0}$-$L_{1}$compaction to reduce write stalls while reducing the depth of LSM-trees to mitigate write amplification. To this end, four novel techniques are proposed. First, we relocate and manage the$L_{0}$level in NVM with our proposedtriangle container. Second, the newright-angle side compactionis devised to compact$L_{0}$to$L_{1}$at fine-grained key ranges, thus substantially reducing the amount of compaction data. Third, TriangleKV increases the width of each level to decrease the depth of LSM-trees thus mitigating write amplification. Finally, thecross-row hint searchis introduced for the triangle container to keep adequate read performance. We implement TriangleKV based on MatrixKV and evaluate it on a hybrid DRAM/NVM/SSD system using Intel's latest 3D Xpoint NVM device Optane DC PMM. Evaluation results show that, with the same amount of NVM, TriangleKV outperforms RocksDB, NoveLSM and MatrixKV in 99th-percentile latencies by$5.5\times$,$2.1\times$and$1.1\times$, and random write throughput by$4.9\times$,$3.5\times$and$1.4\times$respectively.