Approximate Nearest Neighbor Search (ANNS) plays a critical role in applications such as search engines, recommender systems, and RAG for LLMs. Vector quantization (VQ), a crucial technique for ANNS, is commonly used to reduce space overhead and accelerate distance computations. However, despite significant research advances, state-of-the-art VQ methods still face challenges in balancing encoding efficiency and quantization accuracy. To address these limitations, we propose a novel VQ method called SAQ. To improve accuracy, SAQ employs a new dimension segmentation technique to strategically partition PCA-projected vectors into segments along their dimensions. By prioritizing leading dimension segments with larger magnitudes, SAQ allocates more bits to high-impact segments, optimizing the use of the available space quota. An efficient dynamic programming algorithm is developed to optimize dimension segmentation and bit allocation, ensuring minimal quantization error. To speed up vector encoding, SAQ devises a code adjustment technique to first quantize each dimension independently and then progressively refine quantized vectors using a coordinate-descent-like approach to avoid exhaustive enumeration. Extensive experiments demonstrate SAQ's superiority over classical methods (e.g., PQ, PCA) and recent state-of-the-art approaches (e.g., LVQ, Extended RabitQ). SAQ achieves up to 80% reduction in quantization error and accelerates encoding speed by over 80× compared to Extended RabitQ.

This paper studies communication-oblivious algorithms under the massively parallel computation (MPC) model. The communication patterns of these algorithms follow a distribution dependent only on the definition of the underlying problem, the problem size N, and the number p of machines, but not on the specific input elements. Our objective is to understand when obliviousness necessitates --- or does not necessitate --- heavier communication compared to the traditional MPC model that does not enforce such a requirement. The first part of our investigation focuses on single-round algorithms. We prove that skew-free hashing, a fundamental problem solvable with load Õ(N/p) (with high probability or w.h.p.\ for short) under the traditional model, demands a load of nearly Ω(N) under communication obliviousness. Intriguingly, we show that hashing can still be applied in an oblivious manner to process any natural join in one round with a load complexity matching that of the best traditional MPC algorithm. The second part of our investigation studies compilation methods that convert a traditional MPC algorithm A into a communication-oblivious counterpart. Given an A that operates within l = poly(p) rounds and entails a load at most L = Ω(p log p) w.h.p., we can produce w.h.p.\ a communication-oblivious version running in 2l rounds with a load at most (1 + δ) L, where δ < 0 can be an arbitrarily small constant. Additionally, we establish hardness results indicating that the theoretical guarantees of our compilation can no longer be significantly improved.

This paper initializes the study of range subgraph counting and range subgraph listing, both of which are motivated by the significant demands in practice to perform graph analytics on subgraphs pertinent to only selected, as opposed to all, vertices. In the first problem, there is an undirected graph G where each vertex carries a real-valued attribute. Given an interval q and a pattern Q, a query counts the number of occurrences of Q in the subgraph of G induced by the vertices whose attributes fall in q. The second problem has the same setup except that a query needs to enumerate (rather than count) those occurrences with a small delay. In both problems, our goal is to understand the tradeoff between space usage and query cost, or more specifically: (i) given a target on query efficiency, how much pre-computed information about G must we store? (ii) Or conversely, given a budget on space usage, what is the best query time we can hope for? We establish a suite of upper- and lower-bound results on such tradeoffs for various query patterns.

We present a dynamic index structure for join sampling. Built for an (equi-) join Q --- let IN be the total number of tuples in the input relations of Q --- the structure uses ~O(IN) space, supports a tuple update of any relation in ~O(1) time, and returns a uniform sample from the join result in ~O(INρ* / /max{1, OUT} ) time with high probability (w.h.p.), where OUT and ρ* are the join's output size and fractional edge covering number, respectively; notation ~O(.) hides a factor polylogarithmic to IN. We further show how our result justifies the O(INρ* ) running time of existing worst-case optimal join algorithms (for full result reporting) even when OUT łl INρ*. Specifically, unless the combinatorial k-clique hypothesis is false, no combinatorial algorithms (i.e., algorithms not relying on fast matrix multiplication) can compute the join result in O(INρ*-ε ) time w.h.p. even if OUT łe INε, regardless of how small the constant ε > 0 is.

Achieving high performance OLAP over large graphs is a challenging problem and has received great attention recently because of its broad spectrum of applications. Existing systems have various performance bottlenecks due to limitations such as low parallelism and high network overheads. This Demo presents Grasper, an RDMA-enabled distributed graph OLAP system, which adopts a series of new system designs to overcome the challenges of OLAP on graphs. The take-aways for Demo attendees are: (1)~a good understanding of the challenges of processing graph OLAP queries; (2)~useful insights about where Grasper's good performance comes from; (3)~inspirations about how to design an efficient graph OLAP system by comparing Grasper with existing systems.

Similarity search is a core component in various applications such as image matching and product recommendation. However, single-machine solutions are usually insufficient due to the large cardinality of modern datasets. We present Pyramid, a general and efficient framework for distributed similarity search. Pyramid supports search with popular similarity functions including Euclidean distance, angular distance and inner product. Different from existing distributed solutions that are based on KD-tree or locality sensitive hashing (LSH), Pyramid is based on the Hierarchical Navigable Small World graph (HNSW), which is the state-of-the-art similarity search algorithm. To achieve high query processing throughput, Pyramid partitions a dataset into sub-datasets containing similar items for index building and assigns a query to only some of the sub-datasets for query processing. Experiments on large-scale datasets show that Pyramid produces quality results for similarity search, achieves high query processing throughput and low latency, and is robust under node failure and straggler.