SQLoop: High Performance Iterative Processing in Data Management

Distributed Graph Processing: Techniques and Systems

“…Maiter [54] proposes delta-based graph computation, which abstracts the graph computation as an update accumulation process and can avoid invalid (zero-delta) updates to improve computation efficiency. SQLoop [10] leverages DBMS to implement graph iterative computations (DBMS-based) and extends standard SQL with efficient recursive aggregation support. Socialite [30], BigDatalog [32], and PowerLog [40] rely on Datalog language to express distributed graph algorithms (Datalog-based) and allow users to use very concise declarative programs to specify large-scale graph computations.…”

Storage and Navigation Operations on Graphs in Relational DBMS

“…Regarding recursive query optimizations, there are approaches for reducing its intermediary size's results (ORDONEZ, 2005); partitioning methods for the edge table (GAO et al, 2014;HONG et al, 2017;SEO;AHN;IM, 2019); specialized data representations based on adjacency lists (SUN et al, 2015;CHEN, 2013;AHMED;THOMO, 2017); in-memory graph process engines (MA et al, 2016;ZHAO et al, 2019); relational matrix algebra adaptations (DOLMATOVA; AUGSTEN; BÖHLEN, YU, 2017). The relevance of recursive queries as a field of study can be seen through several research articles (ORDONEZ; CABRERA; GURRAM, 2017;JINDAL et al, 2015;RAJ;PATEL, 2015;SIMMEN et al, 2014;PASSING et al, 2017;FLORATOS et al, 2018;ORDONEZ, 2017). Hence, the large quantity of research and studies over recursive queries motivated us to further improve recursive queries' performance.…”

“…In SQL, recursive queries are primarily expressed employing Common Table Expressions (CTEs). CTEs have been part of the SQL standard since 1999 (FLORATOS et al, 2018), and it is supported by most of the RDBMSs such as Microsoft SQL Server, PostgreSQL, SQLite, and Oracle. However, it is not currently available in a range of columnar DBMSs (ORDONEZ; CABRERA; GURRAM, 2017), such as MonetDB, Vertica, and C-Store.…”

“…The quantity of join operations performed in a CTE resulting table corresponds to the max path size found so far in the graph (see Section 2.2.2). A recursive query is one of the most challenging query classes, and it demanded several studies to improve its execution performance (GAO et al, 2011;GAO et al, 2014;RAJ;PATEL, 2015;JINDAL et al, 2015;ORDONEZ, 2005;WISNIEWSKI, 2009;ORDONEZ, 2010;ORDONEZ;RAI, 2014;FLORATOS et al, 2018;YU, 2017). These queries are important in the context of this Ph.D. research since they enable executing, for example, the following graph queries through SQL commands in a RDBMS (CABRERA; ORDONEZ, 2016; CABRERA; ORDONEZ, 2017): Reachability, SSSP (using the Bellman-Ford algorithm), Weakly Connected Components (WCC), and PageRank.…”

“…Lastly, the SQLoop (FLORATOS et al, 2018) also enforces the duplicate row elimination rule by proposing the ITERATIVE operation. The ITERATIVE operation replaces the RECURSIVE one by updating the values instead of only appending them.…”