B4: Experience with a Globally-Deployed Software Defined WAN
Proceedings of the ACM SIGCOMM 2013 conference on SIGCOMM, 2013 · paper · Computer science
Presents the design, implementation and evaluation of B4, Google's private wide-area network between data centers, built on software-defined networking with OpenFlow and centralized traffic engineering. It drives links to near 100% utilization, with lessons from three years of production.
Checked: PDF from cseweb.ucsd.edu, fetched 4 Oct 2026. Page numbers are those of that PDF.
0 findings
None that passed a second check.
9 minor (typos and the like; most checked by the judge only)
Reference or label Figure 4: the RIB lists "1.2.5.0/24 → intf2, intf3" but the derived Flows table lists "1.2.5.0/25 : out to 37"; the prefix length should match (/24), as it does for 1.2.3.0/24.
- p. 4 Figure 4, RIB vs Flows cache
1.2.5.0/24 → intf2, intf3 ... 1.2.5.0/25 : out to 37
Judge's note: Page 4 image: the RIB cylinder shows 1.2.3.0/24 and 1.2.5.0/24; the Flows cylinder shows '1.2.3.0/24 : out to 35' and '1.2.5.0/25 : out to 37'. The text says RAP translates each RIB entry into a Flow table entry, so the /25 looks like a figure typo. The figure text is small, but the digit reads as 5 at page resolution. It is an illustrative example with no effect on results.
Second check: another model agreed
Report a problem with this finding
Wording "all links to average 70% utilization over long time periods, corresponding to 2-3x efficiency improvements relative to standard practice" — against the paper's own baseline of "30-40% average utilization", 70% is 1.75x to 2.33x, which never reaches 3x.
- p. 1 Section 1, right column, TE paragraph
links to average 70% utilization over long time periods, correspond-
- p. 1 Section 1, left column
to 30-40% average utilization
Judge's note: Read page 1 image: both sentences are as quoted. Section 2 also says provisioning at 30-40% is '2-3x the cost of a fully-utilized WAN', so the 2-3x figure seems to echo the near-100% comparison (2.5-3.3x) rather than the 70% average. It is a loose round claim, not a hard numerical error, so I rate it a wording observation.
Second check: another model agreed
Report a problem with this finding
Typo Figure 2 box label reads "Cental TE Server"; it should read "Central TE Server" (the text has "a central TE server").
- p. 3 Figure 2, global layer box
Cental TE Server
Judge's note: Page 3 image: the global-layer box in Fig. 2 reads 'Cental TE Server' with no 'r'. The label is small but legible at page resolution. The body text on the same page uses 'central TE server'.
Second check: none (judge only)
Report a problem with this finding
Typo "the Open Flow protocol to take advantage of the hardware pipeline" — written as two words; the paper uses "OpenFlow" everywhere else, including the same paragraph.
- p. 4 Section 3.2, OFA paragraph
the Open Flow protocol to take advantage of the hardware pipeline
Judge's note: Page 4 image, Section 3.2: 'Open Flow' is printed mid-line with a clear space, so it is not a line-break artefact. 'OpenFlow Agent' and 'OpenFlow (OF)' appear in adjacent lines.
Second check: none (judge only)
Report a problem with this finding
Typo "a Flow table which maps prefixes to entries into a ECMP Group table" — should be "an ECMP Group table".
- p. 4 Section 3.4, fourth paragraph
a Flow table which maps prefixes to entries into a ECMP Group table
Judge's note: Page 4 image, Section 3.4: the line break falls after 'into a' and the next line begins with 'ECMP Group table'. Article error only; the meaning is unchanged.
Second check: none (judge only)
Report a problem with this finding
Typo "descapsulating-site switches to recognize which tunnel-IDs should be terminated" — "descapsulating" should be "decapsulating".
- p. 6 Section 5.1, second paragraph
descapsulating-site switches to recognize which tunnel-IDs should
- p. 6 Section 5.1, second paragraph
descapsulating-site switches
- p. 6 Section 5.1, last paragraph
descapsulating-site switches to recognize which tunnel-IDs should be terminated
- p. 6 Sec 5.1
descapsulating-site
Judge's note: Page 6 image, Section 5.1: 'descapsulating-site' is printed whole at the start of a line. The same section uses 'a decapsulating switch' a few lines earlier.
Second check: none (judge only)
Report a problem with this finding
Typo "OSFP/IS-IS/iBGP" — "OSFP" should be "OSPF".
- p. 11 Section 8, MPLS-TE paragraph
OSFP/IS-IS/iBGP
Judge's note: Page 11 image, Section 8, MPLS-TE paragraph: printed 'OSFP/IS-IS/iBGP to distribute the state'. This is a transposition of the routing protocol acronym OSPF. The paper does not spell the acronym anywhere else, so the correct form comes from general knowledge of the protocol name.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [10]: "(London, UK, UK, 2002)" — "UK" is duplicated in the place of publication.
- p. 12 Reference [10]
(London, UK, UK, 2002)
Judge's note: Page 12 image, reference [10] (Choi et al.): printed '(London, UK, UK, 2002), ICOIN '02, Springer-Verlag, pp. 699–711'. Duplicated word.
Second check: none (judge only)
Report a problem with this finding
Typo Reference [34] prints the last author's surname as "Akyildiz, I. F., and UhI, G." with a capital I as the final letter; the surname is presumably "Uhl".
- p. 12 Reference [34]
Akyildiz, I. F., and UhI,
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: that the fifth author of the TEAM paper (Comm. Mag. 42, 2004) is named Uhl. On the crop the author names are set in small caps; a lowercase l would appear as a small-cap L with a foot, but the final glyph is a plain vertical stroke (I), and the text layer also gives a capital I. So the page does print "UhI" and this is not an extraction artefact. The correct spelling cannot be checked from the files.
Second check: none (judge only)