skip to main content


The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 5:00 PM ET until 11:00 PM ET on Friday, June 21 due to maintenance. We apologize for the inconvenience.

Title: SafeBound: A Practical System for Generating Cardinality Bounds
Recent work has reemphasized the importance of cardinality estimates for query optimization. While new techniques have continuously improved in accuracy over time, they still generally allow for under-estimates which often lead optimizers to make overly optimistic decisions. This can be very costly for expensive queries. An alternative approach to estimation is cardinality bounding, also called pessimistic cardinality estimation, where the cardinality estimator provides guaranteed upper bounds of the true cardinality. By never underestimating, this approach allows the optimizer to avoid potentially inefficient plans. However, existing pessimistic cardinality estimators are not yet practical: they use very limited statistics on the data, and cannot handle predicates. In this paper, we introduce SafeBound, the first practical system for generating cardinality bounds. SafeBound builds on a recent theoretical work that uses degree sequences on join attributes to compute cardinality bounds, extends this framework with predicates, introduces a practical compression method for the degree sequences, and implements an efficient inference algorithm. Across four workloads, SafeBound achieves up to 80% lower end-to-end runtimes than PostgreSQL, and is on par or better than state of the art ML-based estimators and pessimistic cardinality estimators, by improving the runtime of the expensive queries. It also saves up to 500x in query planning time, and uses up to 6.8x less space compared to state of the art cardinality estimation methods.  more » « less
Award ID(s):
1907997 2109922
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Proceedings of the ACM on Management of Data
Page Range / eLocation ID:
1 to 26
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. null (Ed.)
    Accurate selectivity estimation for string predicates is a long-standing research challenge in databases. Supporting pattern matching on strings (such as prefix, substring, and suffix) makes this problem much more challenging, thereby necessitating a dedicated study. Traditional approaches often build pruned summary data structures such as tries followed by selectivity estimation using statistical correlations. However, this produces insufficiently accurate cardinality estimates resulting in the selection of sub-optimal plans by the query optimizer. Recently proposed deep learning based approaches leverage techniques from natural language processing such as embeddings to encode the strings and use it to train a model. While this is an improvement over traditional approaches, there is a large scope for improvement. We propose Astrid, a framework for string selectivity estimation that synthesizes ideas from traditional and deep learning based approaches. We make two complementary contributions. First, we propose an embedding algorithm that is query-type (prefix, substring, and suffix) and selectivity aware. Consider three strings 'ab', 'abc' and 'abd' whose prefix frequencies are 1000, 800 and 100 respectively. Our approach would ensure that the embedding for 'ab' is closer to 'abc' than 'abd'. Second, we describe how neural language models could be used for selectivity estimation. While they work well for prefix queries, their performance for substring queries is sub-optimal. We modify the objective function of the neural language model so that it could be used for estimating selectivities of pattern matching queries. We also propose a novel and efficient algorithm for optimizing the new objective function. We conduct extensive experiments over benchmark datasets and show that our proposed approaches achieve state-of-the-art results. 
    more » « less
  2. Recently there has been significant interest in using machine learning to improve the accuracy of cardinality estimation. This work has focused on improving average estimation error, but not all estimates matter equally for downstream tasks like query optimization. Since learned models inevitably make mistakes, the goal should be to improve the estimates that make the biggest difference to an optimizer. We introduce a new loss function, Flow-Loss, for learning cardinality estimation models. Flow-Loss approximates the optimizer's cost model and search algorithm with analytical functions, which it uses to optimize explicitly for better query plans. At the heart of Flow-Loss is a reduction of query optimization to a flow routing problem on a certain "plan graph", in which different paths correspond to different query plans. To evaluate our approach, we introduce the Cardinality Estimation Benchmark (CEB) which contains the ground truth cardinalities for sub-plans of over 16 K queries from 21 templates with up to 15 joins. We show that across different architectures and databases, a model trained with Flow-Loss improves the plan costs and query runtimes despite having worse estimation accuracy than a model trained with Q-Error. When the test set queries closely match the training queries, models trained with both loss functions perform well. However, the Q-Error-trained model degrades significantly when evaluated on slightly different queries (e.g., similar but unseen query templates), while the Flow-Loss-trained model generalizes better to such situations, achieving 4 -- 8× better 99th percentile runtimes on unseen templates with the same model architecture and training data. 
    more » « less
  3. null (Ed.)
    Query Optimization remains an open problem for Big Data Management Systems. Traditional optimizers are cost-based and use statistical estimates of intermediate result cardinalities to assign costs and pick the best plan. However, such estimates tend to become less accurate because of filtering conditions caused either from undetected correlations between multiple predicates local to a single dataset, predicates with query parameters, or predicates involving user-defined functions (UDFs). Consequently, traditional query optimizers tend to ignore or miscalculate those settings, thus leading to suboptimal execution plans. Given the volume of today’s data, a suboptimal plan can quickly become very inefficient. In this work, we revisit the old idea of runtime dynamic optimization and adapt it to a shared-nothing distributed database system, AsterixDB. The optimization runs in stages (re-optimization points), starting by first executing all predicates local to a single dataset. The intermediate result created from each stage is used to re-optimize the remaining query. This re-optimization approach avoids inaccurate intermediate result cardinality estimations, thus leading to much better execution plans. While it introduces the overhead for materializing these intermediate results, our experiments show that this overhead is relatively small and it is an acceptable price to pay given the optimization benefits. In fact, our experimental evaluation shows that runtime dynamic optimization leads to much better execution plans as compared to the current default AsterixDB plans as well as to plans produced by static cost-based optimization (i.e. based on the initial dataset statistics) and other state-of-the-art approaches. 
    more » « less
  4. Large volumes of videos are continuously recorded from cameras deployed for traffic control and surveillance with the goal of answering “after the fact” queries: identify video frames with objects of certain classes (cars, bags) from many days of recorded video. Current systems for processing such queries on large video datasets incur either high cost at video ingest time or high latency at query time. We present Focus, a system providing both low-cost and low-latency querying on large video datasets. Focus’s architecture flexibly and effectively divides the query processing work between ingest time and query time. At ingest time (on live videos), Focus uses cheap convolutional network classifiers (CNNs) to construct an approximate index of all possible object classes in each frame (to handle queries for any class in the future). At query time, Focus leverages this approximate index to provide low latency, but compensates for the lower accuracy of the cheap CNNs through the judicious use of an expensive CNN. Experiments on commercial video streams show that Focus is 48× (up to 92×) cheaper than using expensive CNNs for ingestion, and provides 125× (up to 607×) lower query latency than a state-of-the-art video querying system (NoScope). 
    more » « less
  5. Modern data systems are typically both complex and general-purpose. They are complex because of the numerous internal knobs and parameters that users need to manually tune in order to achieve good performance; they are general-purpose because they are designed to handle diverse use cases, and therefore often do not achieve the best possible performance for any specific use case. A recent trend aims to tackle these pitfalls: instance-optimized systems are designed to automatically self-adjust in order to achieve the best performance for a specific use case, i.e., a dataset and query workload. Thus far, the research community has focused on creating instance-optimized database components, such as learned indexes and learned cardinality estimators, which are evaluated in isolation. However, to the best of our knowledge, there is no complete data system built with instance-optimization as a foundational design principle. In this paper, we present a progress report on SageDB, our effort towards building the first instance-optimized data system. SageDB synthesizes various instance-optimization techniques to automatically specialize for a given use case, while simultaneously exposing a simple user interface that places minimal technical burden on the user. Our prototype outperforms a commercial cloud-based analytics system by up to 3X on end-to-end query workloads and up to 250X on individual queries. SageDB is an ongoing research effort, and we highlight our lessons learned and key directions for future work. 
    more » « less