Apache Hive: A Warehousing Solution over a Map-Reduce Framework

• p. 1 Introduction, para 2

  Hive-Metastore, containing schemas and statistics, which is useful in data exploration and query optimization

• p. 3 Section 3.1, Table bullet

  this facility can be used to store table statistics in the future

Only the authors can say what is right here.

Judge's note: Checked p1 (crop C001-a) and p3 (crop C001-b). The introduction describes statistics as something the metastore already holds and uses for optimization. Section 3.1 presents storing statistics as future work. Section 5 (p4) says the optimizer is a 'naïve rule-based optimizer' and that a cost-based optimizer is still planned. A grep finds no other place where statistics are described as present. The two passages read inconsistently, and only the authors can say which description is meant.

Second check: another model agreed

• p. 2 Section 2.3, second HiveQL query

  DISTRIBUTE BY school, meme

• p. 2 Section 2.3, second HiveQL query

  SORT BY school, meme, cnt desc

• p. 2 Section 2.3 text

  the top 10 memes per school can then be computed by a simple custom Python reduce script top10.py

Only the authors can say what is right here.

Judge's note: Rests on outside knowledge: the semantics of HiveQL DISTRIBUTE BY (hash-partitions rows on the listed columns) and SORT BY (sorts within each reducer). The paper does not define these clauses. The p2 image shows the query exactly as quoted. The subquery groups by (school, meme), so each (school, meme) pair appears only once. Partitioning on (school, meme) can therefore send one school's memes to different reducers, and the trailing 'cnt desc' sort key never changes the order. As printed, the query does not obviously deliver a per-school top 10. The paper calls the example 'highly simplified', so this is put to the authors as a query rather than treated as a definite error.

Second check: another model agreed

• p. 2 Section 2.3

  can be done using HiveQLs map-reduce constructs

• p. 2 Section 2.3

  using HiveQLs map-reduce constructs

• p. 2 Section 2.3

  HiveQLs map-reduce constructs

Judge's note: The p2 image confirms the printed text has no apostrophe, so this is not an extraction artefact.

Second check: none (judge only)

• p. 3 Section 3.1

  traditional warehousing solution (ala Oracle or DB2)

Judge's note: The crop C004-a confirms 'ala' is printed. This informal spelling is common in technical writing, so it is cosmetic only.

Second check: none (judge only)

• p. 4 References, [1]

  A. Pavlo et. al.

Judge's note: The p4 image confirms 'et. al.' in [1] and 'et al.' in [2].

Second check: none (judge only)

• p. 4 Section 3.2, Optimizer bullet (continued)

  the amount of data transfered between operators

Judge's note: The p4 image, top of the left column, confirms the spelling.

Second check: none (judge only)

• p. 4 Section 5, second paragraph

  This demonstration show cases the current ca-

Judge's note: The p4 image shows two separate words inside the line, not split across a line break. Section 4 spells it 'showcase'.

Second check: none (judge only)

• p. 4 Section 5, first paragraph

  ecution engines has forced us to revisit techniques

Judge's note: The p4 image (end of the left column running into the right column of Section 5) confirms the wording.

Second check: none (judge only)

• p. 4 Section 3.2 end

  writes to two tem-

Judge's note: The crop C011-a confirms the wording. The meaning is clear, and the sentence is only awkward ('writes two temporary files to HDFS'), so this is cosmetic.

Second check: none (judge only)