Large-scale matrix computations have become indispensable in artificial intelligence and scientific applications. It is of paramount importance to efficiently perform out-of-core computations that often entail an excessive amount of disk I/O. Unfortunately, however, most existing systems do not focus on disk I/O aspects and are vulnerable to performance degradation when the scale of input matrices and intermediate data grows large. To address this problem, we present a new out-of-core matrix computation system called PreVision. The PreVision system can achieve optimal buffer replacement by leveraging the deterministic characteristics of data access patterns, and it can also avoid redundant I/O operations by proactively evicting the pages that are no longer referenced. Through extensive evaluations, we demonstrate that PreVision outperforms the existing out-of-core matrix computation systems and significantly reduces disk I/O operations.

The density-based clustering is utilized for various applications such as hot spot detection or segmentation. To serve those applications in real time, it is desired to update clusters incrementally by capturing only the recent data. The previous incremental density-based clustering algorithms often represent clusters as a graph and suffer serious performance degradation. This is because a costly graph traversal is required to check whether a cluster is still connected whenever a point is removed. In order to address the problem of slow deletion, this paper proposes a novel incremental density-based clustering algorithm called DenForest. By maintaining clusters as a group of spanning trees instead of a graph, DenForest can determine efficiently and accurately whether a cluster is to be split by a point removed from the window in logarithmic time. With extensive evaluations, it is demonstrated that DenForest outperforms the state-of-the-art density-based clustering algorithms significantly and achieves the clustering quality comparable with that of DBSCAN.

As the world becomes increasingly data-centric, the tasks dealt with by a database management system (DBMS) become more complex and diverse. Compared with traditional workloads that typically require only a single data model, modern-day computational tasks often involve multiple data sources and rely on more than one data model. Unfortunately, however, there is currently no standard benchmark program that can evaluate a DBMS in the various aspects of multi-model databases, especially when the array data model is concerned. In this paper, we propose M2Bench , a new benchmark program capable of evaluating a multi-model DBMS that supports several important data models such as relational, document-oriented, property graph, and array models. M2Bench consists of multi-model workloads that are inspired by real-world problems. Each task of the workload mimics a real-life scenario where at least two different models of data are involved. To demonstrate the efficacy of M2Bench , we evaluated polyglot or multi-model database systems with the M2Bench workloads and unfolded the diverse characteristics of the database systems for each data model.

Given the prevalence of mobile and IoT devices, continuous clustering against streaming data has become an essential tool of increasing importance for data analytics. Among many clustering approaches, the density-based clustering has garnered much attention due to its unique advantages. The main drawback is, however, the limited scalability attributed to its relatively high computational cost, which is further aggravated when it has to update clusters continuously along with evolving data. In this paper, we present a new incremental density-based clustering algorithm called DISC optimized for the sliding window model. DISC is capable of producing exactly the same clustering results as existing methods such as Incremental DBSCAN for streaming data much more quickly and efficiently.

With increasing volumes of scientific data, a scalable and parallel computing framework is required for scientific analysis in computer simulations and experiments. Scientific data are commonly generated in multi-dimensional arrays, and the array data model is appropriate to store them for analysis, including for data mining and arithmetic computation. In this paper, we introduce an array processing system called Spangle. It is implemented on top of Apache Spark, a popular map-reduce framework for complex computation workloads. To support array data computation, we extended Resilient Distributed Dataset (RDD) based on the array data model named ArrayRDD. ArrayRDD is an inherently parallel data structure that provides fault-tolerance. In addition, by adopting the array data model, Spangle provides an interface for expressing machine learning algorithms, which heavily rely on linear algebra. We tailored two popular algorithms, PageRank and Stochastic Gradient Descent, for large-scale datasets in Spangle.

Recently there has been an increasing interest in analyzing scientific data generated by observations and scientific experiments. For managing these data efficiently, SciDB, a multi-dimensional array-based DBMS, is suggested. When SciDB processes a query with where predicates, it uses filter operator internally to produce a result array that matches the predicates. Most queries for scientific data analysis utilize spatial information. However, filter operator of SciDB reads all data without considering features of array-based DBMSs and spatial information. In this demo, we present an efficient query processing scheme utilizing characteristics of array-based data, implemented by employing coordinates. It uses a selective scan that retrieves data corresponding to a range that satisfies specific conditions. In our experiments, the selective scan is up to 30x faster than the original scan. We demonstrate that our implementation of the filter operator will reduce the processing time of a selection query significantly and enable SciDB to handle a massive amount of scientific data in more scalable manner.