k-way edge based partitioning algorithms for processing large streaming graphs, such as social networks and web crawls, assign each arriving edge to one of the k partitions. This can result in vertices being replicated on multiple partitions. Typically, such partitioning algorithms aim to balance the edge counts across partitions while minimizing the vertex replication. However, such objectives ignore the community structure inherently embedded in the graph, which is an important quality metric for clustering and graph mining applications that subsequently operate on the partitions. To address this gap, we propose a novel optimization goal to maximize the number of local triangles in the partitions as an additional objective. Triangle count is an effective metric to measure the conservation of community structure. Further, we propose TriParts a family of heuristics for online partitioning over an edge stream. They use three complementary state data structures: Bloom Filters, Triangle Map and High degree Map. Each state adds tangible value to meet our objectives. We validate TriParts on six diverse real world graphs with up to 1.6B edges and varying triangle densities. Our best heuristic outperforms the state-of-the-art DBH and HDRF streaming graph partitioners on the triangle-count metric by up to 4–8.3x while maintaining competitive vertex replication factor and edge-balancing. We achieve an ingest rate of 500k edges/sec on a 16 node cluster. We also offer detailed results on the configuration parameters, scalability and overheads of TriParts, and its practical benefits for distributed graph analytics.

Algorithms for temporal property graphs may be time-dependent (TD), navigating the structure and time concurrently, or time-independent (TI), operating separately on different snapshots. Currently, there is no unified and scalable programming abstraction to design TI and TD algorithms over large temporal graphs. We propose an interval-centric computing model (ICM) for distributed and iterative processing of temporal graphs, where a vertex's time-interval is a unit of data-parallel computation. It introduces a unique time-warp operator for temporal partitioning and grouping of messages that hides the complexity of designing temporal algorithms, while avoiding redundancy in user logic calls and messages sent. GRAPHITE is our implementation of ICM over Apache Giraph, and we use it to design 12 TI and TD algorithms from literature. We rigorously evaluate its performance for diverse real-world temporal graphs - as large as 131M vertices and 5.5B edges, and as long as 219 snapshots. Our comparison with 4 baseline platforms on a 10-node commodity cluster shows that ICM shares compute and messaging across intervals to out-perform them by up to 25×, and matches them even in worst-case scenarios. GRAPHITE also exhibits weak-scaling with near-perfect efficiency.

Component-centric distributed graph processing models that use bulk synchronous parallel (BSP) execution have grown popular. These overcome short-comings of Big Data platforms like Hadoop for processing large graphs. However, literature on formal analysis of these component-centric abstractions for different graphs, graph partitioning, and graph algorithms is lacking. Here, we propose an coarse-grained analytical approach based on a meta-graph sketch to examine the characteristics of component-centric graph programming models. We apply this sketch to subgraph- and block-centric abstractions, and draw a comparison with vertex-centric models like Google's Pregel. We explore the impact of various graph partitioning techniques on the meta-graph, and the impact of the meta-graph on graph algorithms. This decouples large unwieldy graphs and their partitioning artifacts from their algorithmic analysis. We evaluate our approach for five spatial and powerlaw graphs, four different partitioning strategies, and for PageRank and Breadth First Search algorithms. We show that this novel analytical technique is simple, scalable and yet gives a reliable estimate of the number of supersteps, and the communication and computational complexities of the algorithms for various graphs.

The performance of prediction models is often based on “abstract metrics” that estimate the model's ability to limit residual errors between the observed and predicted values. However, meaningful evaluation and selection of prediction models for end-user domains requires holistic and application-sensitive performance measures. Inspired by energy consumption prediction models used in the emerging “big data” domain of Smart Power Grids, we propose a suite of performance measures to rationally compare models along the dimensions of scale independence, reliability, volatility and cost. We include both application independent and dependent measures, the latter parameterized to allow customization by domain experts to fit their scenario. While our measures are generalizable to other domains, we offer an empirical analysis using real energy use data for three Smart Grid applications: planning, customer education and demand response, which are relevant for energy sustainability. Our results underscore the value of the proposed measures to offer a deeper insight into models' behavior and their impact on real applications, which benefit both data mining researchers and practitioners.

Time series datasets are a canonical form of high velocity Big Data, and often generated by pervasive sensors, such as found in smart infrastructure. Performing predictive analytics on time series data can be computationally complex, and requires approximation techniques. In this paper, we motivate this problem using a real application from the smart grid domain. We propose an incremental clustering technique, along with a novel affinity score for determining cluster similarity, which help reduce the prediction error for cumulative time series within a cluster. We evaluate this technique, along with optimizations, using real datasets from smart meters, totaling ~700,000 data points, and show the efficacy of our techniques in improving the prediction error of time series data within polynomial time.

Emerging Big Data applications in areas like ecommerce and energy industry require both online and on-demand queries to be performed over vast and fast data arriving as streams. These present novel challenges to Big Data management systems. Complex Event Processing (CEP) is recognized as a high performance online query scheme which in particular deals with the velocity aspect of the 3-V's of Big Data. However, traditional CEP systems do not consider data variety and lack the capability to embed ad hoc queries over the volume of data streams. In this paper, we propose H2O, a stateful complex event processing framework, to support hybrid online and on-demand queries over realtime data. We propose a semantically enriched event and query model to address data variety. A formal query algebra is developed to precisely capture the stateful and containment semantics of online and on-demand queries. We describe techniques to achieve the interactive query processing over realtime data featured by efficient online querying, dynamic stream data persistence and on-demand access. The system architecture is presented and the current implementation status reported.