GreyCat: Efficient what-if analytics for data in motion at scale

Abstract

Over the last few years, data analytics shifted from a descriptive era, confined to the explanation of past events, to the emergence of predictive techniques. Nonetheless, existing predictive techniques still fail to effectively explore alternative futures, which continuously diverge from current situations when exploring the effects of what-if decisions. Enabling prescriptive analytics therefore calls for the design of scalable systems that can cope with the complexity and the diversity of underlying data models. In this article, we address this challenge by combining graphs and time series within a scalable storage system that can organize a massive amount of unstructured and continuously changing data into multi-dimensional data models, called Many-Worlds Graphs. We demonstrate that our open source implementation, GreyCat, can efficiently fork and update thousands of parallel worlds composed of millions of timestamped nodes, such as what-if exploration.

Towards WCET Estimation of Graph Queries@Run.time

“…While the models@run.time initiative [1] has been promoting the use of models, queries and transformations at runtime with major recent advances [2]- [5], existing approaches provide no timeliness guarantees required for any critical applications. For example, while [3] claims "near real-time" for online data analytics, a timely response is not guaranteed. Similarly, the query-based runtime monitoring approach [2] is unable to ensure that a query will justifiably complete on time even in the absence of communication errors.…”

“…In particular, we present a research agenda to assess worst case execution time for graph-based query techniques used at runtime over dynamically evolving graph-like runtime data and a heterogeneous computing platform with resource constraints that is characteristic to many critical CPS applications. Existing applications of such queries@run.time already include online analytics [3] or runtime monitoring [2].…”

“…Since many IoT and CPS applications rely upon graphlike knowledge representation, our concepts are generally applicable to different domains. Nevertheless, we illustrate these concepts in the context of runtime monitoring in an open-source CPS educational demonstrator of a model railway network, which is a representative application of using graph models and graph queries at runtime [2], [3]. A. Runtime Monitoring by Graph Queries Recently, various expressive rule or query-based specification languages have been proposed [2], [7], [8] to specify monitoring goals and to drive monitor execution over a continuously changing runtime graph model following the models@run.time principle [1].…”

Towards an Efficient Approach to Manage Graph Data Evolution: Conceptual Modelling and Experimental Assessments

“…They can evolve over time in terms of (i) their topology (how entities are linked, when entities/relationships are present or absent), (ii) their inherent features (the attributes set that describes an entity or a relationship) and (iii) their status (the values of the set of descriptive attributes at a particular time). Finding and analyzing these evolutions enable to get a deeper understanding of an application notably to exploit temporal correlations and causality [9,3], to make simulations [8] or to make predictions [16]. It is therefore necessary to be able to manage the temporal evolution of data to exploit them.…”

“…d is an ALLEN temporal operator to express that a time interval X occurs "during" a time interval Y, i.e. XdY [1] 8. • is an ALLEN temporal operator to express that a time interval X "overlaps" a time interval Y, i.e.…”