MADlib is a free, open-source library of in-database analytic methods. It provides an evolving suite of SQL-based algorithms for machine learning, data mining and statistics that run at scale within a database engine, with no need for data import/export to other tools. The goal is for MADlib to eventually serve a role for scalable database systems that is similar to the CRAN library for R: a community repository of statistical methods, this time written with scale and parallelism in mind.In this paper we introduce the MADlib project, including the background that led to its beginnings, and the motivation for its opensource nature. We provide an overview of the library's architecture and design patterns, and provide a description of various statistical methods in that context. We include performance and speedup results of a core design pattern from one of those methods over the Greenplum parallel DBMS on a modest-sized test cluster. We then report on two initial efforts at incorporating academic research into MADlib, which is one of the project's goals.MADlib is freely available at http://madlib.net, and the project is open for contributions of both new methods, and ports to additional database platforms.

The increasing use of statistical data analysis in enterprise applications has created an arms race among database vendors to offer ever more sophisticated in-database analytics. One challenge in this race is that each new statistical technique must be implemented from scratch in the RDBMS, which leads to a lengthy and complex development process. We argue that the root cause for this overhead is the lack of a unified architecture for in-database analytics. Our main contribution in this work is to take a step towards such a unified architecture. A key benefit of our unified architecture is that performance optimizations for analytics techniques can be studied generically instead of an ad hoc, per-technique fashion. In particular, our technical contributions are theoretical and empirical studies of two key factors that we found impact performance: the order data is stored, and parallelization of computations on a single-node multicore RDBMS. We demonstrate the feasibility of our architecture by integrating several popular analytics techniques into two commercial and one open-source RDBMS. Our architecture requires changes to only a few dozen lines of code to integrate a new statistical technique. We then compare our approach with the native analytics tools offered by the commercial RDBMSes on various analytics tasks, and validate that our approach achieves competitive or higher performance, while still achieving the same quality.

Statistical information extraction (IE) programs are increasingly used to build real-world IE systems such as Alibaba, CiteSeer, Kylin, and YAGO. Current statistical IE approaches consider the text corpora underlying the extraction program to be static. However, many real-world text corpora are dynamic (documents are inserted, modified, and removed). As the corpus evolves, and IE programs must be applied repeatedly to consecutive corpus snapshots to keep extracted information up to date. Applying IE from scratch to each snapshot may be inefficient: a pair of consecutive snapshots may change very little, but unaware of this, the program must run again from scratch. In this paper, we present CRFlex, a system that efficiently executes such repeated statistical IE, by recycling previous IE results to enable incremental update. As the first step, CRFlex focuses on statistical IE programs which use a leading statistical model, Conditional Random Fields (CRFs). We show how to model properties of the CRF inference algorithms for incremental update and how to exploit them to correctly recycle previous inference results. Then we show how to efficiently capture and store intermediate results of IE programs for subsequent recycling. We find that there is a tradeoff between the I/O cost spent on reading and writing intermediate results, and CPU cost we can save from recycling those intermediate results. Therefore we present a cost-based solution to determine the most efficient recycling approach for any given CRF-based IE program and an evolving corpus. We conduct extensive experiments with CRF-based IE programs for 3 IE tasks over a real-world data set to demonstrate the utility of our approach.

No abstract

No abstract