CockroachDB: The Resilient Geo-Distributed SQL Database
Proceedings - ACM-SIGMOD International Conference on Management of Data, 2020 · paper · Computer science
Presents the design of CockroachDB, a scalable SQL database for global transactional workloads, with a transaction model for consistent geo-distributed transactions. It covers replication, the distributed SQL layer, performance evaluation, case studies and lessons from five years of development.
Checked: PDF from cockroachlabs.com, fetched 5 Oct 2026. Page numbers are those of that PDF.
1 finding
Substance Table 1, CockroachDB 100,000-warehouse column prints Max tpmC "1,245,462" with Efficiency "98.8%"; on the per-warehouse basis that reproduces the other four efficiency cells this gives about 96.8%.
- p. 11 Table 1, CockroachDB block, 100,000-warehouse column
1,245,462
- p. 11 Table 1, CockroachDB Efficiency row, 100,000-warehouse column
98.8%
- p. 11 Section 6.1.3, text citing Table 1
at near-maximum efficiency
- p. 11 Table 1, CockroachDB, 100,000 warehouses column
Efficiency 97.0% 96.5% 98.8%
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the TPC-C maximum of 12.86 tpmC per warehouse, which the paper does not state. The table itself supports that constant: the other four cells recompute to 97.0%, 96.45%, 97.8% and 7.3%, as printed. Only the 100,000-warehouse cell is off: 1,245,462/1,286,000 = 96.85% against the printed 98.8%. No footnote or text explains a different basis for that column. Either the tpmC or the efficiency is misprinted, and only the authors can say which. Values read from the table crop and the full page image.
Second check: another model agreed
Report a problem with this finding
9 minor (typos and the like; most checked by the judge only)
Wording Algorithm 1 line 11 tests only "if op.key unchanged over (txnTimestamp, resp.ts]", while the prose says these lines verify that "repeating the previous reads in the transaction at the new timestamp will return the same value (Lines 11 and 12)" and Section 3.4 says the refresh re-scans the read set.
- p. 4 Algorithm 1, line 11
if op.key unchanged over (txnTimestamp, resp.ts]
- p. 5 Section 3.1.1, first paragraph on the page
peating the previous reads in the transaction at the new timestamp will return the same value (Lines 11 and 12)
- p. 6 Section 3.4, third paragraph
validates that the keys have not been updated in a given timestamp interval (Algorithm 1, Lines 11 to 14)
Judge's note: I read line 11 in the crop and the two prose passages on pages 5 and 6. The pseudocode names only the current operation's key; the prose attributes a check of the whole read set to the same lines. The algorithm is presented as high-level and may be deliberately simplified, so this is a minor consistency point.
Second check: another model agreed
Report a problem with this finding
Wording Section 3.3.2 says "a read on the same key at a higher timestamp tb >= ta": the word "higher" is strict while the printed condition allows equality.
- p. 6 Section 3.3.2 Read-write conflicts
same key at a higher timestamp tb >= ta
Judge's note: The crop shows "higher timestamp t_b >= t_a" as quoted. The word and the symbol do not state the same condition, though "higher or equal" is plausibly what is meant. Minor wording imprecision.
Second check: another model agreed
Report a problem with this finding
Typo Algorithm 1, line 17 prints "asynchonously notify leaseholder to commit"; the word should be "asynchronously".
- p. 4 Algorithm 1 (Transaction Coordinator), line 17
asynchonously notify leaseholder to commit
Judge's note: The crop of Algorithm 1 shows the spelling "asynchonously" (no 'r' after 'h') on line 17, matching the text layer. It is a plain misspelling.
Second check: none (judge only)
Report a problem with this finding
Typo Algorithm 2 line 9 reads "if not op.commit then", the only conditional in Algorithms 1 and 2 that carries a trailing "then".
- p. 5 Algorithm 2 (Leaseholder), line 9
if not op.commit then
Judge's note: The crop shows a bold "then" on line 9 only. The other conditionals visible in Algorithm 2 (lines 5, 11, 14) and Algorithm 1 (lines 4, 10, 11, 16) have none. Cosmetic inconsistency; the meaning is unchanged.
Second check: none (judge only)
Report a problem with this finding
Typo Section 6.3 prints "For most YSCB workloads, CRDB shows significantly higher throughput"; the benchmark is YCSB.
- p. 12 Section 6.3, second paragraph, first sentence
For most YSCB workloads
- p. 12 Section 6.3, second paragraph, first sentence
For most YSCB workloads, CRDB shows significantly higher
Judge's note: The crop shows "YSCB" in this sentence and "YCSB" a few lines below in the same paragraph. The text layer has YCSB in footnote 2 and reference [76]. Letter transposition.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [26] title prints "Volcano/spl minus/an extensible and parallel query evaluation system", with unrendered markup "/spl minus/" in place of a dash.
- p. 15 Reference [26]
Volcano/spl minus/an extensible and parallel
Judge's note: The crop of the reference list shows the literal string "/spl minus/" in the title of [26]. It is leftover symbol markup from bibliographic metadata.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [23] prints the publisher inside stray quotation marks: “ O’Reilly Media, Inc.”.
- p. 15 Reference [23]
O’Reilly Media, Inc.
Judge's note: The crop shows opening and closing quotation marks around "O'Reilly Media, Inc." in [23]. Other publishers in the list (for example Pearson Education in [6]) are unquoted. Bibliography-source leftover.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [47] has a doubled full stop: "Self-Driving Database Management Systems.. In CIDR".
- p. 16 Reference [47]
Self-Driving Database Management Systems.. In CIDR
Judge's note: The crop shows two full stops after "Systems" in [47].
Second check: none (judge only)
Report a problem with this finding
Typo Reference [77] has a doubled full stop: "Global-Scale Placement of Transactional Data Stores.. In EDBT".
- p. 17 Reference [77]
Global-Scale Placement of Transactional Data Stores.. In EDBT
Judge's note: The crop shows two full stops after "Stores" in [77].
Second check: none (judge only)