The Snowflake Elastic Data Warehouse

• p. 7 Section 4.3, third paragraph

  Snowflake handles parsing and type inference (cf. Section 4.3.3)

• p. 8 Section 4.3.2

  to perform automatic type inference and to determine which (typed) paths are frequently common

• p. 8

  which (typed) paths are frequently common

Judge's note: On page 7 the pointer reads 4.3.3. On page 8, Section 4.3.2 introduces "a novel automated approach to type inference and columnar storage" and describes "automatic type inference". Section 4.3.3 covers string-to-type conversion at write or read time. That is related, so the pointer is imprecise rather than plainly wrong, which is why this is an observation. 'frequently common' is a minor redundancy (S4) in the same passage.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 2 Section 1, 'Highly Available' bullet

  Snowflake tolerates node, cluster, and even full data center failures.

• p. 6 Section 4.2.1, last paragraph

  we today accept this one scenario of partial system unavailability

• p. 6 Section 4.2.1, last paragraph

  the user needs to actively re-provision the VW in a different AZ

Only the authors can say what is right here.

Judge's note: I checked both passages on the page images (p2 Highly Available bullet; p6 Section 4.2.1, last paragraph). The storage and Cloud Services layers may well survive an AZ loss, so the intro claim is not false for the whole system. Still, the summary claim is unqualified while the body acknowledges partial unavailability, which needs manual action, for exactly this failure. This is an overstatement in wording, not a factual error.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 9 Section 4.3.4, results paragraph

  query processing was around 10% for all but two queries (Q9 and Q17 over SF1000)

• p. 8 Figure 4, Q8 bars in both panels

  Figure 4: TPC-H SF100 and SF1000 Performance: Relational vs. Schema-less Row Format

• p. 9 Section 4.3.4, results paragraph

  the overhead of schema-less storage and query processing was around 10% for all but two queries (Q9 and Q17 over SF1000)

Only the authors can say what is right here.

Judge's note: I read the Figure 4 bars on page 8 (approximate). Q8 SF100: relational ≈4.1 s vs schema-less ≈5.5 s. Q8 SF1000: ≈30 vs ≈40. SF1000 Q19: ≈19 vs ≈25. SF100 Q5: ≈4.0 vs ≈4.6. Q9 and Q17 at SF1000 are indeed the largest outliers. However, Q8 in particular is well above 'around 10%' in both panels and is not named as an exception. The figure readings are approximate, so this is put as a query on wording.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 10 Section 4.5.3 End-to-End Security

  connected to a virtual private cluster within AWS

Only the authors can say what is right here.

Judge's note: Rests on outside knowledge: AWS CloudHSM instances are deployed into a Virtual Private Cloud (VPC), and AWS has no 'virtual private cluster' construct. The page 10 image (Section 4.5.3) shows 'virtual private cluster' as quoted.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 4 Section 3.2.1, second paragraph

  use cases where performance isolation is not big concern.

• p. 4 Section 3.2.1, second paragraph

  where performance isolation is not big concern

Judge's note: The page 4 image (Section 3.2.1, second paragraph) shows "is not big concern.)". It is a real typo, not an extraction artefact.

Second check: none (judge only)

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 7 Section 4.2.2, second paragraph

  at every which point all new queries issued

• p. 7 Section 4.2.2, second paragraph

  new version, at every which point all new queries issued

Judge's note: The page 7 image (Section 4.2.2, second paragraph) shows "at every which point" exactly as quoted. It is a copy-editing slip.

Second check: none (judge only)

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 9 Section 4.4, first SQL example

  ’Mon, 01 May 2015 16:20:00 -0700’::timestamp);

• p. 9 Section 4.4, first SQL example

  'Mon, 01 May 2015 16:20:00 -0700'::timestamp

Only the authors can say what is right here.

Judge's note: Rests on outside knowledge: 1 January 2015 was a Thursday, which makes 1 May 2015 a Friday. The page 9 image shows 'Mon, 01 May 2015'. Days from 1 Jan to 1 May = 31+28+31+30 = 120, and 120 mod 7 = 1, so the date is one weekday after Thursday. This is a cosmetic inconsistency in an illustrative example.

Second check: none (judge only)

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 12 Section 6, first paragraph on page

  both for standard SQL and semi-relational extensions

• p. 11 Section 6, third paragraph

  we had hoped that the semi-structured extensions would prove useful

Judge's note: The page 12 image shows 'semi-relational extensions'. A text search finds 'semi-relational' only here, while 'semi-structured extensions' is used on page 11 and 'semi-structured' throughout. This is probably a slip in terminology, but the intended term cannot be certain.

Second check: none (judge only)

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send