Algebraic Optimization of RDF Graph Pattern Queries on MapReduce - Large Scale and Big Data: Processing and Management - page 223

(a)

Varying multiplicity of MV property
Pig-Opt

3500

Hive NTGA

3000

2500

2000

1500

1000

500

0

low-1Star high-1Star base-2Star ow-2Starhigh-2Star

(b)

Redundancy factor across the MapReduce workflow

Query MR S 1 MR S 1 S 2

MR S 2

Low-1Star 0.72 (1.5 GB) - -

High-1Star 0.82 (5.2 GB) - -

Base-2Star 0 (0.2 GB) 0 (7.7 GB) 0 (12.7 GB)

Low-2Star 0.72 (1.6 GB) 0 (7.7 GB) 0.78 (75.8 GB)

High-2Star 0.82 (5.4 GB) 0 (7.7 GB) 0.89 (250 GB)

(c)

Varying density of star-joins

2500

Pig-Opt Hive NTGA-Opt

2000

1500

1000

500

0

MV-2p MV-3p MV-4p V-5p

FIGURE 6.20 (a) Comparative evaluation using one and two star subpattern queries con-
taining low and high multiplicity MV property, (b) redundancy factor in reduce output while
evaluating test queries using flat algebra, (c) impact of lazy unnesting strategy with increasing
cardinality of star-joins (BSBM-500k, 10-node).

6.10 CONCLUDING REMARKS
This chapter discusses the challenges and strategies for RDF query processing
on MapReduce platforms. The impetus for this research direction is the range of
emerging applications that rely on increasing amounts of publicly available Semantic
Web data as background knowledge for analysis. In many scenarios, the computa-
tional needs required to incorporate such large amounts of Semantic Web data in