B-Trees are the predominant index structures in relational databases. However, they become a performance bottleneck in traversal-heavy workloads. Although learned indexes have emerged as a promising solution to mitigate this bottleneck, their high engineering complexity hinders their practical adoption as replacements for B-Trees. Instead, we propose a more viable alternative: augmenting B-Trees with learned indexing techniques to reduce integration costs and speed up traversals. To this end, we present MirrorBtr, a unified framework that builds upon existing learned indexes to accelerate B-Tree traversals without altering the B-Tree structure. MirrorBtr reduces space and synchronization overhead and dynamically adapts to B-Tree updates. We implement MirrorBtr in MySQL InnoDB and conduct extensive evaluations, showing that it consistently delivers traversal speedups with negligible maintenance overhead and outperforms both the conventional B-Tree and prior acceleration methods in InnoDB. In particular, MirrorBtr achieves up to 7.45× higher query throughput than the B-Tree.
Intersecting sets of integers is a fundamental operation in search engines and graph analytics. Yet most approaches treat integers as opaque values and compare them one by one, ignoring that every integer set partitions the key space into occupied and empty ranges. PRISM exploits this: it decomposes 64-bit integers into fixed-width prefix segments organized in a trie, where each node uses a bitmap to record which sub-ranges are occupied. Ranges absent from at least one operand are pruned without examining individual elements, fixed-width leaf segments map directly onto SIMD registers, and the structure extends naturally to k -way intersection via simultaneous descent. Morphing-node layouts adjust each node's size by occupancy, reducing memory usage by up to 8× on sparse data while minimally affecting throughput on dense data. We also evaluate PRISM on three real-world case studies: the Star Schema Benchmark, a TPC-H trigram index, and inverted list intersection on ClueWeb09. On all three, PRISM is the fastest method for intersection-heavy workloads, ahead of compressed-bitmap methods and more than an order of magnitude faster than general-purpose ordered and hash-based containers.
The growing prevalence of multimodal data streams-such as video, images, and text-poses new challenges for stream processing systems. Existing systems are designed for structured data and struggle to support multimodal workloads efficiently. While multimodal large language models (MLLMs) provide powerful semantic understanding, naively integrating them into streaming pipelines can incur excessive computation and degrade performance. This demonstration presents SAM(center dot) S & amacr;RA, a novel super-optimizer that enables real-time processing of multimodal data streams by aggressively specializing query plans to the query and data stream. Through an interactive traffic monitoring scenario, the demo demonstrates how multimodal queries are optimized by transforming naive query plans into optimized query plans. Users can directly observe how semantic, logical, and physical optimizations interact, reduce redundant model invocations, and impact system behavior, including responsiveness and execution accuracy. The demonstration shows that efficient multimodal stream processing requires rethinking query optimization beyond naive MLLM integration.
For decades, database software has reacted to hardware. AI-for-Systems is about to shorten the development time gap between them. This panel asks the DB community what genuine HW-SW co-design should look like in the coming decade, and whether we are organized to do it.
Formulating effective filtering conditions for database queries often requires a deep understanding of the specific values stored within table columns, which can impose a significant cognitive burden on users, especially when the textual columns contain hundreds or even thousands of unique values. To address this challenge, we introduce FilterPilot, an LLM-powered interactive assistant designed to adapt filtering predicates to table content. Given user-provided filtering terms and optional time constraints, FilterPilot assists by suggesting appropriate filtering conditions. At its core, FilterPilot employs a novel iterative recall-then-verify paradigm, combining LLM-based query reformulation with table-value feedback to dynamically expand search terms. This process is further optimized through an Upper Confidence Bound (UCB) selection mechanism and an early stopping strategy. We release the source code, datasets, and a live demo system on https://github.com/wusw14/FilterPilot to encourage community adoption.
