Lazy-Adaptive Tree

Abstract

Flash memories are in ubiquitous use for storage on sensor nodes, mobile devices, and enterprise servers. However, they present significant challenges in designing tree indexes due to their fundamentally different read and write characteristics in comparison to magnetic disks. In this paper, we present the Lazy-Adaptive Tree (LA-Tree), a novel index structure that is designed to improve performance by minimizing accesses to flash. The LA-tree has three key features: 1) it amortizes the cost of node reads and writes by performing update operations in a lazy manner using cascaded buffers, 2) it dynamically adapts buffer sizes to workload using an online algorithm, which we prove to be optimal under the cost model for raw NAND flashes, and 3) it optimizes index parameters, memory management, and storage reclamation to address flash constraints. Our performance results on raw NAND flashes show that the LA-Tree achieves 2x to 12x gains over the best of alternate schemes across a range of workloads and memory constraints. Initial results on SSDs are also promising, with 3x to 6x gains in most cases.

An Efficient Indexing Structure for Multidimensional Categorical Range Aggregation Query

“…Collecting the cache not only takes up memory but also has a negative impact on query performance. Therefore, some scholars have proposed LA-Tree [17], which can better adapt to different loads than Buffer-Tree. As mentioned earlier, this idea has also been adopted in the aBB-Tree to reduce the update overhead.…”

A New Flash-based B+-Tree with Very Cheap Update Operations on Leaf Nodes

“…Since the B+-tree has a majority of updates within its leaf level and the leaf nodes seems to be scattered across storage, a flash B+-tree seems to inevitably suffer from frequent full merges in the presence of very frequent key updates. To solve such an inherent problem of the B+-tree stored in flash, some ideas are proposed to decrease full merges during the garbage collection time [4,12,13,15]. The flash B+-trees proposed in [13,15] take the same approach in that key updates are not reflected in real-time on the data of leaf nodes.…”

“…To solve such an inherent problem of the B+-tree stored in flash, some ideas are proposed to decrease full merges during the garbage collection time [4,12,13,15]. The flash B+-trees proposed in [13,15] take the same approach in that key updates are not reflected in real-time on the data of leaf nodes. Instead, the corresponding log data is recorded in other tree levels above leaf level.…”

“…Moreover, as the price per bit is recently decreasing at a very fast rate, there is a growing demand for flash-based DBMSs that are able to successfully process harsh workloads of enterprise-scale transaction processing system [3,5,9]. In this light, many researches have been conducted to develop a flash-aware B+-tree that can index efficiently a large volume of records on flash storage such as the solid state drive (SSD) [4,[11][12][13][14][15][16].…”

“…Because the B-tree index was invented for its use in HDDs, there is a concern that the B-tree index may suffer from poor stability or performance degradation because of any different I/O feature of flash memory from HDDs [11][12][13][14][15][16]. As a matter of fact, the absence of in-place updating in flash's functionalities has been regarded as a major technical disadvantage to be tackled [1,12,13].…”