Efficient execution plans for distributed skyline query processing

Abstract

In this paper, we study the generation of efficient execution plans for skyline query processing in large-scale distributed environments. In such a setting, each server stores autonomously a fraction of the data, thus all servers need to process the skyline query. An execution plan defines the order in which the individual skyline queries are processed on different servers, and influences the performance of query processing. Querying servers consecutively reduces the amount of transferred data and the number of queried servers, since skyline points obtained by one server prune points in the subsequent servers, but also increases the latency of the system. To address this trade-off, we introduce a novel framework, called SkyPlan, for processing distributed skyline queries that generates execution plans aiming at optimizing the performance of query processing. Thus, we quantify the gain of querying consecutively different servers. Then, execution plans are generated that maximize the overall gain, while also taking into account additional objectives, such as bounding the maximum number of hops required for the query or balancing the load on different servers fairly. Finally, we present an algorithm for distributed processing based on the generated plan that continuously refines the execution plan during in-network processing. Our framework consistently outperforms the state-of-the-art algorithm.

A survey of skyline processing in highly distributed environments

“…SkyPlan was proposed in [36] for improving the performance of PaDSkyline [13]. Similar to PaDSkyline, during query processing, each peer reports a set of minimum bounding rectangles (MBRs) to the querying peer as a summarization of its data.…”

“…On the other hand, AGiDS requires global knowledge of a grid structure with identical boundaries for all peers, which may not be feasible in some applications. Furthermore, in [36] it has been shown that SkyPlan outperforms PaDSkyline due to a more efficient execution plan. In case multiple hops are required for contacting each peer, the cost of these approaches increases since at most N p hops are required for gathering the summaries or querying the peers at each step leading to a cost of O(N p )+ O(N p ) (for AGiDS) or O(N p )+ O(2· N p ) (for PaDSkyline, SkyPlan).…”

“…Approaches that use MBRs for query routing, such as PaDSkyline [13], SkyPlan [36], and DDS [18,19], can easily be adapted to support subspace skyline queries since the MBRs can be projected into the given subspace before being processed. Then, peers can be pruned based on the projected MBRs, and the correctness of the approaches is still guaranteed.…”

“…Therefore, it is not guaranteed that they contain at least one skyline point. SkyPlan [36], similar to PaDSkyline, can efficiently support constrained skyline queries by discarding MBRs that do not overlap with the given constraints.…”

“…For example, [50] exploits the applicability of distributed indexing techniques proposed for P2P systems for cloud infrastructures. In fact, some of the approaches [8,13,35,36,56] described in this survey were proposed for a more general distributed architecture than P2P and therefore they can easily be adapted to other distributed systems. Moreover, although this paper focuses on highly distributed environments, we also present briefly skyline query processing in other distributed systems (web information systems and wireless sensor networks) as well as different data types (stream data and uncertain data).…”