Search

Estimating the selectivity of a query is a key step in almost any cost-based query optimizer. Most of today's databases rely on histograms or samples that are periodically refreshed by re-scanning the data as the underlying data changes. Since frequent scans are costly, these statistics are often stale and lead to poor selectivity estimates. As an alternative to scans, query-driven histograms have been proposed, which refine the histograms based on the actual selectivities of the observed queries. Unfortunately, these approaches are either too costly to use in practice---i.e., require an exponential number of buckets---or quickly lose their advantage as they observe more queries. In this paper, we propose a selectivity learning framework, called QuickSel, which falls into the query-driven paradigm but does not use histograms. Instead, it builds an internal model of the underlying data, which can be refined significantly faster (e.g., only 1.9 milliseconds for 300 queries). This fast refinement allows QuickSel to continuously learn from each query and yield increasingly more accurate selectivity estimates over time. Unlike query-driven histograms, QuickSel relies on a mixture model and a new optimization algorithm for training its model. Our extensive experiments on two real-world datasets confirm that, given the same target accuracy, QuickSel is 34.0x--179.4x faster than state-of-the-art query-driven histograms, including ISOMER and STHoles. Further, given the same space budget, QuickSel is 26.8%--91.8% more accurate than periodically-updated histograms and samples, respectively.

Despite decades of research on approximate query processing (AQP), our understanding of sample-based joins has remained limited and, to some extent, even superficial. The common belief in the community is that joining random samples is futile. This belief is largely based on an early result showing that the join of two uniform samples is not an independent sample of the original join, and that it leads to quadratically fewer output tuples. However, unfortunately, this result has little applicability to the key questions practitioners face. For example, the success metric is often the final approximation's accuracy, rather than output cardinality. Moreover, there are many non-uniform sampling strategies that one can employ. Is sampling for joins still futile in all of these settings? If not, what is the best sampling strategy in each case? To the best of our knowledge, there is no formal study answering these questions. This paper aims to improve our understanding of sample-based joins and offer a guideline for practitioners building and using real-world AQP systems. We study limitations of offline samples in approximating join queries: given an offline sampling budget, how well can one approximate the join of two tables? We answer this question for two success metrics: output size and estimator variance. We show that maximizing output size is easy, while there is an information-theoretical lower bound on the lowest variance achievable by any sampling strategy. We then define a hybrid sampling scheme that captures all combinations of stratified, universe, and Bernoulli sampling, and show that this scheme with our optimal parameters achieves the theoretical lower bound within a constant factor. Since computing these optimal parameters requires shuffling statistics across the network, we also propose a decentralized variant where each node acts autonomously using minimal statistics., 19 pages

We demonstrate VerdictDB, the first platform-independent approximate query processing (AQP) system. Unlike existing AQP systems that are tightly-integrated into a specific database, VerdictDB operates at the driver-level, acting as a middleware between users and off-the-shelf database systems. In other words, VerdictDB requires no modifications to the database internals; it simply relies on rewriting incoming queries such that the standard execution of the rewritten queries under relational semantics yields approximate answers to the original queries. VerdictDB exploits a novel technique for error estimation called variational subsampling, which is amenable to efficient computation via SQL. In this demonstration, we showcase VerdictDB's performance benefits (up to two orders of magnitude) compared to the queries that are issued directly to existing query engines. We also illustrate that the approximate answers returned by VerdictDB are nearly identical to the exact answers. We use Apache Spark SQL and Amazon Redshift as two examples of modern distributed query platforms. We allow the audience to explore VerdictDB using a web-based interface (e.g., Hue or Apache Zeppelin) to issue queries and visualize their answers. VerdictDB is currently open-sourced and available under Apache License (V2).

Lock managers are a crucial component of modern distributed systems. However, with the increasing availability of fast RDMA-enabled networks, traditional lock managers can no longer keep up with the latency and throughput requirements of modern systems. Centralized lock managers can ensure fairness and prevent starvation using global knowledge of the system, but are themselves single points of contention and failure. Consequently, they fall short in leveraging the full potential of RDMA networks. On the other hand, decentralized (RDMA-based) lock managers either completely sacrifice global knowledge to achieve higher throughput at the risk of starvation and higher tail latencies, or they resort to costly communications in order to maintain global knowledge, which can result in significantly lower throughput. In this paper, we show that it is possible for a lock manager to be fully decentralized and yet exchange the partial knowledge necessary for preventing starvation and thereby reducing tail latencies. Our main observation is that we can design a lock manager primarily using RDMA's fetch-and-add (FA) operations, which always succeed, rather than compare-and-swap (CAS) operations, which only succeed if a given condition is satisfied. While this requires us to rethink the locking mechanism from the ground up, it enables us to sidestep the performance drawbacks of the previous CAS-based proposals that relied solely on blind retries upon lock conflicts. Specifically, we present DSLR (Decentralized and Starvation-free Lock management with RDMA), a decentralized lock manager that targets distributed systems running on RDMA-enabled networks. We demonstrate that, despite being fully decentralized, DSLR prevents starvation and blind retries by guaranteeing first-come-first-serve (FCFS) scheduling without maintaining explicit queues. We adapt Lamport's bakery algorithm to an RDMA-enabled environment with multiple bakers, utilizing only one-sided READ and atomic FA operations. Our experiments show that, on average, DSLR delivers 1.8x (and up to 2.8x) higher throughput than all existing RDMA-based lock managers, while reducing their mean and 99.9% latencies by 2.0x and 18.3x (and up to 2.5x and 47x), respectively.

Despite 25 years of research in academia, approximate query processing (AQP) has had little industrial adoption. One of the major causes of this slow adoption is the reluctance of traditional vendors to make radical changes to their legacy codebases, and the preoccupation of newer vendors (e.g., SQL-on-Hadoop products) with implementing standard features. Additionally, the few AQP engines that are available are each tied to a specific platform and require users to completely abandon their existing databases---an unrealistic expectation given the infancy of the AQP technology. Therefore, we argue that a universal solution is needed: a database-agnostic approximation engine that will widen the reach of this emerging technology across various platforms. Our proposal, called VerdictDB, uses a middleware architecture that requires no changes to the backend database, and thus, can work with all off-the-shelf engines. Operating at the driver-level, VerdictDB intercepts analytical queries issued to the database and rewrites them into another query that, if executed by any standard relational engine, will yield sufficient information for computing an approximate answer. VerdictDB uses the returned result set to compute an approximate answer and error estimates, which are then passed on to the user or application. However, lack of access to the query execution layer introduces significant challenges in terms of generality, correctness, and efficiency. This paper shows how VerdictDB overcomes these challenges and delivers up to 171$\times$ speedup (18.45$\times$ on average) for a variety of existing engines, such as Impala, Spark SQL, and Amazon Redshift, while incurring less than 2.6% relative error. VerdictDB is open-sourced under Apache License., Extended technical report of the paper that appeared in Proceedings of the 2018 International Conference on Management of Data, pp. 1461-1476. ACM, 2018

The rising volume of datasets has made training machine learning (ML) models a major computational cost in the enterprise. Given the iterative nature of model and parameter tuning, many analysts use a small sample of their entire data during their initial stage of analysis to make quick decisions (e.g., what features or hyperparameters to use) and use the entire dataset only in later stages (i.e., when they have converged to a specific model). This sampling, however, is performed in an ad-hoc fashion. Most practitioners cannot precisely capture the effect of sampling on the quality of their model, and eventually on their decision-making process during the tuning phase. Moreover, without systematic support for sampling operators, many optimizations and reuse opportunities are lost. In this paper, we introduce BlinkML, a system for fast, quality-guaranteed ML training. BlinkML allows users to make error-computation tradeoffs: instead of training a model on their full data (i.e., full model), BlinkML can quickly train an approximate model with quality guarantees using a sample. The quality guarantees ensure that, with high probability, the approximate model makes the same predictions as the full model. BlinkML currently supports any ML model that relies on maximum likelihood estimation (MLE), which includes Generalized Linear Models (e.g., linear regression, logistic regression, max entropy classifier, Poisson regression) as well as PPCA (Probabilistic Principal Component Analysis). Our experiments show that BlinkML can speed up the training of large-scale ML tasks by 6.26x-629x while guaranteeing the same predictions, with 95% probability, as the full model., Comment: 22 pages, SIGMOD 2019

There has been substantial research on sub-linear time approximate algorithms for Maximum Inner Product Search (MIPS). To achieve fast query time, state-of-the-art techniques require significant preprocessing, which can be a burden when the number of subsequent queries is not sufficiently large to amortize the cost. Furthermore, existing methods do not have the ability to directly control the suboptimality of their approximate results with theoretical guarantees. In this paper, we propose the first approximate algorithm for MIPS that does not require any preprocessing, and allows users to control and bound the suboptimality of the results. We cast MIPS as a Best Arm Identification problem, and introduce a new bandit setting that can fully exploit the special structure of MIPS. Our approach outperforms state-of-the-art methods on both synthetic and real-world datasets., Comment: AAAI 2019

Recent years have witnessed a surge of interest in Approximate Query Processing (AQP) solutions, both in academia and the commercial world. In addition to well-known open problems in this area, there are many new research challenges that have surfaced as a result of the first interaction of AQP technology with commercial and real-world customers. We categorize these into deployment, planning, and interface challenges. At the same time, AQP settings introduce many interesting opportunities that would not be possible in a database with precise answers. These opportunities create hopes for overcoming some of the major limitations of traditional database systems. For example, we discuss how a database can reuse its past work in a generic way, and become smarter as it answers new queries. Our goal in this talk is to suggest some of the exciting research directions in this field that are worth pursuing.

While much of the research on transaction processing has focused on improving overall performance in terms of throughput and mean latency, surprisingly less attention has been given to performance predictability: how often individual transactions exhibit execution latency far from the mean. Performance predictability is increasingly important when transactions lie on the critical path of latency-sensitive applications, enterprise software, or interactive web services. In this paper, we focus on understanding and mitigating the sources of performance unpredictability in today's transactional databases. We conduct the first quantitative study of major sources of variance in MySQL, Postgres (two of the largest and most popular open-source products on the market), and VoltDB (a non-conventional database). We carry out our study with a tool called TProfiler that, given the source code of a database system and programmer annotations indicating the start and end of a transaction, is able to identify the dominant sources of variance in transaction latency. Based on our findings, we investigate alternative algorithms, implementations, and tuning strategies to reduce latency variance without compromising mean latency or throughput. Most notably, we propose a new lock scheduling algorithm, called Variance-Aware Transaction Scheduling (VATS), and a lazy buffer pool replacement policy. In particular, our modified MySQL exhibits significantly lower variance and 99th percentile latencies by up to 5.6× and 6.3×, respectively. Our proposal has been welcomed by the open-source community, and our VATS algorithm has already been adopted as of MySQL's 5.7.17 release (and been made the default scheduling policy in MariaDB).

In today's databases, previous query answers rarely benefit answering future queries. For the first time, to the best of our knowledge, we change this paradigm in an approximate query processing (AQP) context. We make the following observation: the answer to each query reveals some degree of knowledge about the answer to another query because their answers stem from the same underlying distribution that has produced the entire dataset. Exploiting and refining this knowledge should allow us to answer queries more analytically, rather than by reading enormous amounts of raw data. Also, processing more queries should continuously enhance our knowledge of the underlying distribution, and hence lead to increasingly faster response times for future queries. We call this novel idea---learning from past query answers---Database Learning. We exploit the principle of maximum entropy to produce answers, which are in expectation guaranteed to be more accurate than existing sample-based approximations. Empowered by this idea, we build a query engine on top of Spark SQL, called Verdict. We conduct extensive experiments on real-world query traces from a large customer of a major database vendor. Our results demonstrate that Verdict supports 73.7% of these queries, speeding them up by up to 23.0x for the same accuracy level compared to existing AQP systems., Comment: This manuscript is an extended report of the work published in ACM SIGMOD conference 2017

In recent years, our customers have expressed frustration in the traditional approach of using a combination of disparate products to handle their streaming, transactional and analytical needs. The common practice of stitching heterogeneous environments in custom ways has caused enormous production woes by increasing development complexity and total cost of ownership. With SnappyData, an open source platform, we propose a unified engine for real-time operational analytics, delivering stream analytics, OLTP and OLAP in a single integrated solution. We realize this platform through a seamless integration of Apache Spark (as a big data computational engine) with GemFire (as an in-memory transactional store with scale-out SQL semantics). In this demonstration, after presenting a few use case scenarios, we exhibit SnappyData as our our in-memory solution for delivering truly interactive analytics (i.e., a couple of seconds), when faced with large data volumes or high velocity streams. We show that SnappyData can exploit state-of-the-art approximate query processing techniques and a variety of data synopses. Finally, we allow the audience to define various high-level accuracy contracts (HAC), to communicate their accuracy requirements with SnappyData in an intuitive fashion.

Running an online transaction processing (OLTP) system is one of the most daunting tasks required of database administrators (DBAs). As businesses rely on OLTP databases to support their mission-critical and real-time applications, poor database performance directly impacts their revenue and user experience. As a result, DBAs constantly monitor, diagnose, and rectify any performance decays. Unfortunately, the manual process of debugging and diagnosing OLTP performance problems is extremely tedious and non-trivial. Rather than being caused by a single slow query, performance problems in OLTP databases are often due to a large number of concurrent and competing transactions adding up to compounded, non-linear effects that are difficult to isolate. Sudden changes in request volume, transactional patterns, network traffic, or data distribution can cause previously abundant resources to become scarce, and the performance to plummet. This paper presents a practical tool for assisting DBAs in quickly and reliably diagnosing performance problems in an OLTP database. By analyzing hundreds of statistics and configurations collected over the lifetime of the system, our algorithm quickly identifies a small set of potential causes and presents them to the DBA. The root-cause established by the DBA is reincorporated into our algorithm as a new causal model to improve future diagnoses. Our experiments show that this algorithm is substantially more accurate than the state-of-the-art algorithm in finding correct explanations.

Interactive visualizations are crucial in ad hoc data exploration and analysis. However, with the growing number of massive datasets, generating visualizations in interactive timescales is increasingly challenging. One approach for improving the speed of the visualization tool is via data reduction in order to reduce the computational overhead, but at a potential cost in visualization accuracy. Common data reduction techniques, such as uniform and stratified sampling, do not exploit the fact that the sampled tuples will be transformed into a visualization for human consumption. We propose a visualization-aware sampling (VAS) that guarantees high quality visualizations with a small subset of the entire dataset. We validate our method when applied to scatter and map plots for three common visualization goals: regression, density estimation, and clustering. The key to our sampling method's success is in choosing tuples which minimize a visualization-inspired loss function. Our user study confirms that optimizing this loss function correlates strongly with user success in using the resulting visualizations. We also show the NP-hardness of our optimization problem and propose an efficient approximation algorithm. Our experiments show that, compared to previous methods, (i) using the same sample size, VAS improves user's success by up to 35% in various visualization tasks, and (ii) VAS can achieve a required visualization quality up to 400 times faster.