Consistency of Continuous Queries in Fog Computing

A fog computing architecture consists of edge nodes that generate and possibly pre-process (sensor) data, fog nodes that do some processing quickly and enable any actuations that may be needed, and cloud nodes that may perform further, detailed, analytics for long-term and archival purposes. The input data consist of tuples. They are processed in batches. The computation to be done on a batch is decomposed into sub-computations to be executed at the different nodes in the fog architecture. Guaranteeing consistency of the executions is very important. Consistency issues arise for sub-computations at individual nodes, the entire computations on individual batches and computations over sequences of input batches; the latter form continuous queries (CQs). CQs consist of one-time executions, called 1-CQs, and are partitioned into CQ-segments. In this paper, the sub-computations are assumed to be executed individually and serially at each node. We study global serializability of the entire computations on the batches. For CQs, we study the consistency of the CQs in terms of the sensitivity of the 1-CQs and the CQ-segments to breaks in the sequence of input batches due to some batches not being processed completely. Our study is also applicable to breaks occurring due to some actuations or external updates to the program states that affect the semantics of the input batches.