Article Preview
TopIntroduction
Materializing views is a widely practiced strategy to improve online analytical processing (OLAP) query performance significantly. Since storing all possible views is not a cost-effective approach, the materialized view selection problem (MVS) focuses on selecting an optimal set of views to materialize. Such views help attain reasonable query processing time within adequate storage space (Harinarayan et al., 1996; Lin & Kuo, 2004).
Making the right information available at the right time to the right user is the critical need of the hour. All of the existing materialized view selection approaches treat each of the queries and views equally. However, in the real world, users have varying importance in an organization and have different query performance expectations (Gosain & Madaan, 2018a; Kimball & Caserta, 2004; Sauter, 2014). Thus, queries hold priority values depending on the users. The prioritized materialized view selection problem (PMVS) (Gosain & Madaan, 2018b) leverages this concept to prioritize cubes based on the query priority and provides a higher chance of selecting the cubes/views required by the top users. Furthermore, the authors proposed a priority and frequency-based cost model (PFBCM) by adding view priority into the fitness function for optimization. It has been proven to select a more suitable set of views within the specified space constraints and increase user satisfaction by achieving better query performance.
The search space for the view selection problem increases exponentially with the dimensionality, making it an NP-hard problem (Harinarayan et al., 1996; Lin & Kuo, 2004). In such cases, deterministic methods exhaust the acceptable time limits for searching, whereas evolutionary algorithms, although they may not find an optimal solution, arrive quickly at a near-optimal solution. Evolutionary optimization algorithms are stochastic search algorithms on a random population that follow some species’ natural evolution or social behavior for an optimal solution search (Sanz et al., 2014, Mirjalili et al., 2014, Karaboga & Basturk, 2007). Evolutionary algorithms have been used aggressively to find optimal results in various domains. Many of them, such as the genetic algorithm (GA), particle swarm optimization (PSO), ant colony optimization (ACO), artificial bee colony (ABC) optimization, and others have been explored to obtain an optimal materialized view set (Arun & Kumar, 2017; Gosain & Heena, 2016; Lin & Kuo, 2004; Loureiro & Belo, 2006; Song & Gao, 2010).
The particle swarm optimization (PSO) algorithm has been utilized to solve PMVS and has produced satisfactory results (Gosain & Madaan, 2018b). Recently, naive evolutionary algorithms such as coral reef optimization (CRO) (Sanz et al., 2014), grey wolf optimizer (GWO) (Mirjalili et al., 2014), artificial bee colony (ABC) (Karaboga & Basturk, 2007), ant colony optimization (ACO) (Dorigo et al.,2006), and cuckoo search (CS) (Yang & Deb, 2009) are rapidly adopted in diverse areas for efficient optimization. These algorithms have shown better results than PSO in various problems, but an algorithm working well for one class of problems might not perform well in another. Therefore, this study paper aims to investigate different evolutionary algorithms for solving PMVS. Performance was assessed based on optimized cost objective, convergence speed, and view selection specific metrics, such as detailed cost savings ratio, percentage of views materialized for high priority queries, and query processing cost. Empirical and statistical analysis to discover algorithms best suited for the PMVS problem is presented.