Naiad: A Timely Dataflow System
Proceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles, 2013 · paper · Computer science
Presents Naiad, a distributed system for data-parallel cyclic dataflow programs, built on a model called timely dataflow. It combines batch throughput, stream latency and iterative computation, and is reported to outperform specialized systems in their target domains.
Checked: PDF from microsoft.com, fetched 4 Oct 2026. Page numbers are those of that PDF.
0 findings
None that passed a second check.
7 minor (typos and the like; most checked by the judge only)
Wording Text says "Figure 6d plots the running times for the two applications" but Figure 6d's y-axis is "Speedup vs. single computer".
- p. 11 §5.4, strong scaling paragraph
Figure 6d plots the running times for the two applications.
- p. 10 Figure 6d, y-axis label
Speedup vs. single computer
Judge's note: Figure 6d on the p.10 image plots speedup on a 0-50 axis and shows no running times. The following sentences in the text discuss speedups (38x, 46x), so only the description of the figure is loose.
Second check: another model agreed
Report a problem with this finding
Wording "The optimizations reduce the volume of protocol traffic by one or two orders of magnitude" whereas Figure 6c shows Local+GlobalAcc roughly three orders below None at 64 computers.
- p. 11 §5.3, continued paragraph
The optimizations reduce the volume of protocol traffic by one or two orders of magnitude
- p. 10 Figure 6c (log y-axis, curves None vs Local+GlobalAcc)
Local+GlobalAcc
Judge's note: Read by eye from the log-scale Figure 6c. At 64 computers None is about 3,000-4,000 MB, LocalAcc and GlobalAcc about 70-80 MB (under two orders), and Local+GlobalAcc about 3 MB (about three orders). At smaller cluster sizes the combined gap is nearer two orders. The values are approximate plot readings, so this is a wording looseness, not a definite error.
Second check: another model agreed
Report a problem with this finding
Wording "constant number of edges (18.2M) and nodes (9.1M) per computer" times 64 computers is about 1.16B edges, but the text says "for a 1.1B edge graph run on 64 computers".
- p. 11 §5.4, weak scaling paragraph
constant number of edges (18.2M) and nodes (9.1M) per computer
- p. 11 §5.4, weak scaling paragraph
for a 1.1B edge graph run on 64 computers
- p. 11 Section 5.4, weak scaling
constant number of edges (18.2M)
Judge's note: Read on the p.11 image. 18.2M x 64 = 1,164.8M, which rounds to 1.2B, not 1.1B. Truncation or a binary-unit convention would give 1.1B, so this is a small rounding inconsistency. Likewise 63/64 of 360 MB is 354.4 MB against the printed 355 MB.
Second check: another model agreed
Report a problem with this finding
Typo Text on p.8 writes "Datacenter TCP [6]" while reference [6] gives the title as "Data Center TCP (DCTCP)".
- p. 8 Section 3.5 Networking
Datacenter TCP [6]
- p. 15 Reference [6]
Data Center TCP (DCTCP)
Judge's note: Both spellings are visible on the page images. This is a cosmetic naming variation; either form could be legitimate in running text.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [6] prints author names as "J. Padhyey, P. Pately, B. Prabhakarz, S. Senguptay, and M. Sridharany", with stray trailing 'y'/'z' characters.
- p. 15 References, entry [6]
J. Pad- hyey, P. Pately, B. Prabhakarz, S. Senguptay, and M. Sridharany
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the correct DCTCP author surnames (Padhye, Patel, Prabhakar, Sengupta, Sridharan). The page image shows the names exactly as quoted, so this is not an extraction artefact. The uniform trailing y/z on five names, with the first three authors clean, is consistent with affiliation symbols fused into the names.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [23] spells a co-author "R. Ranakrishnan" while reference [25] spells "R. Ramakrishnan".
- p. 16 References, entry [23]
R. Ranakrishnan
- p. 16 References, entry [25]
R. Ra- makrishnan
Judge's note: Both spellings are visible on the p.16 image, in adjacent Loo et al. entries from the same line of work. The inconsistency is internal to the reference list; "Ranakrishnan" in [23] is the evident misspelling.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [44] prints "A. D.Joseph" with no space between the initial and the surname.
- p. 17 References, entry [44]
A. D.Joseph
Judge's note: The crop of p.17 shows "A. D.Joseph" with no space, unlike the other author names in the list. This is a typesetting slip, not an extraction artefact.
Second check: none (judge only)