Query optimization for OLAP-XML federations

Cubing Algorithms for XML Data

“…A pushforward based heuristics finally retrieves the global answer to Q T . Cost model and estimation of SQL XM query trees are provided by the same authors in [25], whereas in [26] the same authors again focus the attention on query optimization techniques for SQL XM query trees. Still in line with this main research direction, [22] studies the problem of enriching SQL XM by novel characteristics that allow us to achieve adaptive features during query and processing tasks over OLAP-XML federations, with the aim of making SQL XM robust with respect to changes that can occur in external XML data sources, and compliant with load balancing issues over external XML data sources.…”

Integrating Data Warehouses with Web Data: A Survey

“…Algebraic query rewriting rules are applied to push as much of the query evaluation toward the OLAP and XML components as possible. The cost-based optimization strategies are based on the cost model described in [60] and a set of the techniques that include in-lining literal XML data values into OLAP predicates, caching, and prefetching [61].…”

Complex Group-By Queries for XML

“…In particular, the focus of [14] is on structural aggregation by node types as opposed to value aggregation. Studies by Fiebig and Moerkotte [7], Pedersen et al [15], and Deutsch et al [5] all consider using query optimization-style rewrite rules for various kinds of grouping. Similarly, [4,12] study techniques for unnesting nested queries with joins and correlated aggregates in the OO and relational context.…”

“…(2) for each node encountered { (3) if the node is not an α, β, or γ node, skip the node (4) if it is an α node { (5) update appropriate counter if count( * ) is specified (6) if node type inversion is involved, update the dummy α node } (7) if it is a β node { (8) if a new β value is encountered (9) create a new set of group nodes with the group-by label (10) otherwise, update appropriate counters if count( * ) is specified } (11) if it is a γ node { (12) if the aggregation is holistic, (13) output the value and the α-node id to the gamma file (14) otherwise { (15) if the parent-id associated with the counter is the same as the parent-id of the current node, invoke domergesiblings() (16) otherwise, invoke domergenonsiblings() } } (17) } /* end-for */ (18) scan through the gamma file, using the α-node ids to form groups (19) use domergenonsiblings() to compute the aggregation for each group (20) put the computed values in the appropriate nodes of the answer tree } Figure 8. Algorithm NGB-Disk shown in Figure 8, there is a separate gamma file pass after all the input has been read.…”

“…If the constraint is already violated, then the corresponding β group is flagged. Lines (10), (15) and (16) check if the group to be updated is a flagged β group. If so, no updating is required.…”