Visual Exploration of Large Multidimensional Data Using Parallel Coordinates on Big Data Infrastructure

Abstract

Abstract:The increase of data collection in various domains calls for an adaptation of methods of visualization to tackle magnitudes exceeding the number of available pixels on screens and challenging interactivity. This growth of datasets size has been supported by the advent of accessible and scalable storage and computing infrastructure. Similarly, visualization systems need perceptual and interactive scalability. We present a complete system, complying with the constraints of aforesaid environment, for visual exploration of large multidimensional data with parallel coordinates. Perceptual scalability is addressed with data abstraction while interactions rely on server-side data-intensive computation and hardware-accelerated rendering on the client-side. The system employs a hybrid computing method to accommodate pre-computing time or space constraints and achieves responsiveness for main parallel coordinates plot interaction tools on billions of records.

HiePaCo: Scalable Hierarchical Exploration in Abstract Parallel Coordinates Under Budget Constraints

“…Abstraction is one solution to address this problem [7] that is widely used for multiple visualization techniques. It consists in the display of visual aggregates instead of single lines [8,9,10,11]. Figure 2b and Figure 2c shows examples of abstract parallel coordinates using per-dimension clustering to aggregate polylines with two different levels of detail (LoD).…”

“…Despite the aggregation, these plots successfully provide an overview similar to a traditional plot and perceptually scale for any size of input data. Since they are based on reduced data, they decrease rendering time [9] and suit client/server architecture by bounding the size of the data transferred between client and server [11]. Sansen et al [11] also leverage aggregation to bound the storage requirements of some precomputed interactions.…”

“…Since they are based on reduced data, they decrease rendering time [9] and suit client/server architecture by bounding the size of the data transferred between client and server [11]. Sansen et al [11] also leverage aggregation to bound the storage requirements of some precomputed interactions. Precomputing interactions reduces the cost of interactive processing as it remove their dependency on linear scans of the data which is particularly interesting when data does not fit in a desktop computer memory.…”

“…Supporting these interactions is essential but strongly increases the number of states of the visualization. Then, straightforwardly precomputing and storing these states as proposed by Sansen et al [11] is no longer efficient. Rübel et al [12] presented a system based on a modern high-performance computing (HPC) platform for addressing processing scalability.…”

“…On one hand, several abstract parallel coordinate representations based on aggregation have been proposed [8,9,10,11,14,15,16], some together with hierarchical exploration interactions [14,15,16]. On the other hand, parallel-coordinate visualization systems targeting large-scale datasets [11,12] have not addressed the implementation of such interaction in a remote-visualization setting. The aim of this paper is to address the scalable implementation of such interaction in a remote-visualization context backed by a distributed platform.…”