Jianting Zhang

Large-scale spatial join query processing in Cloud

The rapidly increasing amount of location data available in many applications has made it desirable to process their large-scale spatial queries in Cloud for performance and scalability. We report our designs and implementations of two prototype systems that are ready for Cloud deployments: SpatialSpark based on Apache Spark and ISP-MC based on Cloudera Impala. Both systems support indexed spatial joins based on point-in-polygon test and point-to-polyline distance computation. Experiments on the pickup locations of ~170 million taxi trips in New York City and ~10 million global species occurrences records have demonstrated both efficiency and scalability using Amazon EC2 clusters.

Speeding up large-scale point-in-polygon test based spatial join on GPUs

Point-in-Polygon (PIP) test is fundamental to spatial databases and GIS. Motivated by the slow response times in joining large-scale point locations with polygons using traditional spatial databases and GIS, we have designed and developed an end-to-end system completely on Graphics Processing Units (GPUs) to associate points with the polygons that they fall within by utilizing massively data parallel computing power of GPUs. The system includes an efficient module to generate point quadrants that have at most K points from large-scale unordered points, a simple grid-file based spatial filtering approach to associate point quadrants and polygons, and, a PIP test module to assign polygons to points in a GPU computing block using both the block and thread level parallelisms. Experiments on joining 170 million points with more than 40 thousand polygons have resulted in a runtime of 11.165 seconds on an Nvidia Quadro 6000 GPU device. In contrast, a baseline serial CPU implementation using state-of-the-art open source GIS packages required 15+ hours to complete. We further discuss several factors and parameters that may affect the system performance.

Parallel spatial query processing on GPUs using R-trees

R-Trees are popular spatial indexing techniques that have been widely adopted in many geospatial applications. As commodity GPUs (Graphics Processing Units) are increasingly becoming available on personal workstations and cluster computers, there are considerable research interests in applying the massive data parallel GPGPU (General Purpose computing on GPUs) technologies to index and query large-scale geospatial data on GPUs using R-Trees. In this study, we aim at evaluating the potentials of accelerating both R-Tree bulk loading and spatial window query processing on GPUs using R-Trees. In addition to designing an efficient data layout schema for R-Trees on GPUs, we have implemented several parallel spatial window query processing techniques on GPUs using both dynamically generated R-Trees constructed on CPUs and bulk loaded R-Trees constructed on GPUs. Extensive experiments using both synthetic and real-world datasets have shown that our GPU based parallel query processing techniques using R-Trees can achieve about 10X speedups on average over 8-core CPU parallel implementations by effectively utilizing large numbers of processors and high memory bandwidth on GPUs.

Towards personal high-performance geospatial computing (HPC-G): perspectives and a case study

Cluster computing, Cloud computing and GPU computing play overlapping and complementary roles in parallel processing of geospatial data within the general HPC framework. The fast increasing hardware capacities of modern personal computers equipped with chip multiprocessor CPUs and massively parallel GPUs have made high performance computing of large-scale geospatial data in a personal computing environment possible. We discuss the framework of Personal HPC-G and compare it with traditional Cluster computing and the newly emerging Cloud computing. We consider Personal HPC-G possesses many favorable features: low initial and operational costs, good support for data management and excellent support for both numeric modeling and interactive visualization. A case study on developing a parallel spatial statistics module for visual explorations on top of Personal HPC-G is subsequently presented.

Indexing large-scale raster geospatial data using massively parallel GPGPU computing

Advances in geospatial technologies have generated large amounts of raster geospatial data. Massively parallel General Purpose Graphics Processing Unit (GPGPU) computing technologies have provided personal computers with tremendous computing capabilities. In this paper, we report our work on fast indexing of large-scale raster geospatial data using GPGPU computing. We have designed a cache conscious quadtree data structure (CCQ-Tree) that is suitable for GPU indexing. A set of algorithms have been developed and integrated to construct CCQ-Trees on GPUs by utilizing multiple pyramid data structures and Z-order based prefix sum. Experiments on multiple 4096*4096 blocks of a global precipitation raster data have shown that CCQ-Tree indexing using a 112-core Nvidia Quadro FX3700 GPU device reduces construction times from around 9.83 seconds to 0.42 seconds (23X speedup).

Data integration and workflow solutions for ecology

The Science Environment for Ecological Knowledge (SEEK) is designed to help ecologists overcome data integration and synthesis challenges. The SEEK environment enables ecologists to efficiently capture, organize, and search for data and analytical processes. We describe SEEK and discuss how it can benefit ecological niche modeling in which biodiversity scientists require access and integration of regional and global data as well as significant analytical resources.

Parallel quadtree coding of large-scale raster geospatial data on GPGPUs

Global remote sensing and large-scale environmental modeling have generated huge amounts of raster geospatial data. While the inherent data parallelism of large-scale raster geospatial data allows straightforward coarse-grained parallelization at the chunk level on CPUs, it is largely unclear how to effectively exploit such data parallelism on massively parallel General Purpose Graphics Processing Units (GPGPUs) that require fine-grained parallelization. In this study, we have developed an efficient spatial data structure called BQ-Tree to code raster geospatial data by exploiting the uniform distributions of quadrants of bitmaps at the bitplanes of a raster. A fine-grained parallelization scheme has been implemented using Nvidia CUDA. Experiments show that the GPGPU implementation is capable of decoding a BQ-Tree encoded 16-bits NASA MODIS geospatial raster with 22,658*15,586 cells in 190 milliseconds, i.e., 1.86 billion cells per second, on an Nvidia C2050 GPU card. The performance achieves a 5.9X speedup when compared with the best dual quadcore CPU implementation and a 36.9X speedup compared with a highly optimized single core CPU implementation.

High-performance quadtree constructions on large-scale geospatial rasters using GPGPU parallel primitives

The increasingly available graphics processing units (GPU) hardware and the emerging general purpose computing on GPU (GPGPU) technologies provide an attractive solution to high-performance geospatial computing. In this study, we have proposed a parallel, primitive-based approach to quadtree construction by transforming a multidimensional geospatial computing problem into chaining a set of generic parallel primitives that are designed for one-dimensional (1D) arrays. The proposed approach is largely data-independent and can be efficiently implemented on GPGPUs. Experiments on 4096*4096 and 16384*16384 raster tiles have shown that the implementation can complete the quadtree constructions in 13.33 ms and 250.75 ms, respectively, on average on an NVidia GPU device. Compared with an optimized serial CPU implementation based on the traditional recursive depth-first search (DFS) tree traversal schema that requires 1191.87 ms on 4096*4096 raster tiles, a significant speedup of nearly 90X has been observed. The performance of the GPU-based implementation also suggests that an indexing rate in the order of more than one billion raster cells per second can be achieved on commodity GPU devices.

U 2 STRA: high-performance data management of ubiquitous urban sensing trajectories on GPGPUs

Volumes of GPS recorded trajectory data in ubiquitous urban sensing applications are increasing fast. Many trajectory queries are both I/O and computing intensive. In this study, we propose to develop the U2STRA prototype system to efficiently manage large-scale GPS trajectory data using General Purpose computing on Graphics Processing Units (GPGPU) technologies. Towards this end, we have developed a trajectory data layout schema using simple in-memory array structures which is not only flexible for data accesses but also cache friendly. We have further developed an end-to-end trajectory similarity query processing technique on GPUs. Our experiments on two publically available large trajectory datasets (GeoLife and T-Drive) have demonstrated the efficiency of massively data parallel GPGPU computing. An impressive 87X speedup for spatial aggregations of GPS point locations and 25-40X speedups for trajectory queries over serial CPU implementations have been achieved. The U2STRA system has also been integrated with commercial desktop and Web-based GIS systems and spatial databases for visual exploration purposes.

Ontology-Driven Composition and Validation of Scientific Grid Workflows in Kepler: a Case Study of Hyperspectral Image Processing

Hyperspectral image processing based on grid computing technology is attractive due to the large data volumes of hyperspectral images and intensive computation requirements for processing. Many existing grid workflow tools do not provide integrated visual workflow composition environments and/or do not have workflow validation mechanisms to ensure structural and semantic correctness of composed grid workflows. In this study, we use Kepler scientific workflow system to compose and validate hyperspectral image processing grid workflows due to its integrated workflow composition environment, rich support for grid computing and built-in infrastructures for structural and semantic workflow validations. We have developed workflow component ontology and data type ontology and plugged them into Kepler. Experiments have been performed to demonstrate the feasibility and effectiveness of the proposed approach