We introduce time curves as a general approach for visualizing patterns of evolution in temporal data. Examples of such patterns include slow and regular progressions, large sudden changes, and reversals to previous states. These patterns can be of interest in a range of domains, such as collaborative document editing, dynamic network analysis, and video analysis. Time curves employ the metaphor of folding a timeline visualization into itself so as to bring similar time points close to each other. This metaphor can be applied to any dataset where a similarity metric between temporal snapshots can be defined, thus it is largely datatype-agnostic. We illustrate how time curves can visually reveal informative patterns in a range of different datasets.

Abstract We introduce MultiPiles, a visualization to explore time‐series of dense, weighted networks. MultiPiles is based on the physical analogy of piling adjacency matrices, each one representing a single temporal snapshot. Common interfaces for visualizing dynamic networks use techniques such as: flipping/animation; small multiples; or summary views in isolation. Our proposed ‘piling’ metaphor presents a hybrid of these techniques, leveraging each one's advantages, as well as offering the ability to scale to networks with hundreds of temporal snapshots. While the MultiPiles technique is applicable to many domains, our prototype was initially designed to help neuroscientists investigate changes in brain connectivity networks over several hundred snapshots. The piling metaphor and associated interaction and visual encodings allowed neuroscientists to explore their data, prior to a statistical analysis. They detected high‐level temporal patterns in individual networks and this helped them to formulate and reject several hypotheses.

Research into social network analysis has shown that graph metrics, such as degree and closeness, are often used to summarize structural changes in a dynamic graph. However there have been few visual analytics approaches that have been proposed to help analysts study graph evolutions in the context of graph metrics. In this paper, we present a novel approach, called GraphFlow, to visualize dynamic graphs. In contrast to previous approaches that provide users with an animated visualization, GraphFlow offers a static flow visualization that summarizes the graph metrics of the entire graph and its evolution over time. Our solution supports the discovery of high-level patterns that are difficult to identify in an animation or in individual static representations. In addition, GraphFlow provides users with a set of interactions to create filtered views. These views allow users to investigate why a particular pattern has occurred. We showcase the versatility of GraphFlow using two different datasets and describe how it can help users gain insights into complex dynamic graphs.

I/O traces are crucial for understanding the performance of new storage architectures. Unfortunately, traces are extremely bursty and difficult to characterize. They are large, difficult to obtain, and unwieldy. In this paper, we examine a trace synthesis method based on cluster analysis of time-varying characteristics of I/O traces. Representative trace segments are selected, and synthetic traces are reconstructed from these segments. We show that we can achieve a 5-10% demerit factor for I/O response times with a reduction of trace data volume of 75-90%.

Accurate disk workloads are crucial for storage systems design, but I/O traces are difficult to obtain, unwieldy to work with, and unparameterizable. I/O traces are often bursty and difficult to characterize. Although good models of I/O workloads would be extremely useful, such bursty traces cannot accurately be modeled using exponential or Poisson arrival times. Much experimental evidence suggests that I/O traces are self-similar, which researchers have hoped might help to model bursty traces. In this paper, we show that self-similarity at large time scales does not significantly affect disk behavior with respect to response times. This allows us to generate synthetic arrival patterns at relatively small time scales, improving the accuracy of trace generation. The relative error of our method, with input parameters suitable for the workload, ranges from approximately 8% to 12%.

Network, Web, and disk I/O traffic are usually bursty and self-similar and therefore cannot be modeled adequately with Poisson arrivals. However, we wish to model these types of traffic and generate realistic traces, because of obvious applications for disk scheduling, network management, and Web server design. Previous models (like fractional Brownian motion and FARIMA, etc.) tried to capture the 'burstiness'. However, the proposed models either require too many parameters to fit and/or require prohibitively large (quadratic) time to generate large traces. We propose a simple, parsimonious method, the b-model, which solves both problems: it requires just one parameter, and can easily generate large traces. In addition, it has many more attractive properties: (a) with our proposed estimation algorithm, it requires just a single pass over the actual trace to estimate b. For example, a one-day-long disk trace in milliseconds contains about 86 Mb data points and requires about 3 minutes for model fitting and 5 minutes for generation. (b) The resulting synthetic traces are very realistic: our experiments on real disk and Web traces show that our synthetic traces match the real ones very well in terms of queuing behavior.

The performance gap between microprocessors and secondary storage is still a limitation in today's systems. Academia and industry are developing new technologies to overcome this gap, such as improved read-write head technology and higher storage densities. One promising new technology is probe-based storage[1]. Characteristics of probe-based storage include small size, high density, high parallelism, low power consumption, and rectilinear motion. We have created a probe-based storage simulation model, configurable to different design points, and identify its sensitivity to various parameters.

Input/output performance on current parallel file systems is sensitive to a good match of application access patterns to file system capabilities. Automatic input/output access pattern classification can determine application access patterns at execution time, guiding adaptive file system policies. In this paper, we examine and compare two novel input/output access pattern classification methods based on learning algorithms. The first approach uses a feedforward neural network previously trained on access pattern benchmarks to generate qualitative classifications. The second approach uses hidden Markov models trained on access patterns from previous executions to create a probabilistic model of input/output accesses. In a parallel application, access patterns can be recognized at the level of each local thread or as the global interleaving of all application threads. Classification of patterns at both levels is important for parallel file system performance; we propose a method for forming global classifications from local classifications. We present results from parallel and sequential benchmarks and applications that demonstrate the viability of this approach.

Optimizing collective input/output (I/O) is important for improving throughput of parallel scientific applications.Current research suggests that a specialized collective application programming interface, coupled with system-level optimizations, is necessary to obtain good I/O performance.Unfortunately, collective interfaces require an application to disclose its entire access pattern to fully reorder I/O requests, and cannot flexibly utilize additional memory to improve performance.In this paper we propose and analyze a method of optimizing collective access patterns using informed prefetching that is capable of exploiting any amount of available memory to overlap I/O with computation.We compare this approach to diskdirected I/O, an efficient implementation of a collective I/O interface.Moreover, we prove that under certain conditions, a per-processor prefetch depth equal to the number of drives can guarantee sequential disk accesses for any collectively accessed file.In empirical studies, a prefetch horizon of one to two times the number of disks per processor is sufficient to match the performance of disk-directed I/O for sequentially allocated files.Finally, we develop accurate analytical models to predict the throughput of informed prefetching for collective reads as a function of the per-processor prefetch depth.

Parallel input/output systems attempt to alleviate the performance bottleneck that affects many input/output intensive applications. In such systems, an understanding of the application access pattern, especially how requests from multiple processors for different file regions are logically related, is important for optimizing file system performance. We propose a method for automatically classifying these global access patterns and using these global classifications to select and tune file system policies to improve input/output performance. We demonstrate this approach on benchmarks and scientific applications using global classification to automatically select appropriate underlying Intel PFS input/output modes and server buffering strategies.