Talk:Google data centers/Archive 1

Servers section
Unsure about the Servers section of the article and how it relates to the Google Filesystem, which employs a hierarchy of Master Server, Replica Servers, Chunk Servers, and eventually clients themselves. --Sgkay 19:21, 28 September 2005 (UTC)

Accuracy
I have an objection to "THz" as a measure of computing power. This seems to be a reference to simply summing the clock speeds of a number of different computers (like saying that 5 machines with 500 MHz clocks will have "2.5 GHz" of processing power). This additive measure is very misleading; why not phrase the calculation in terms of a unit that is meaningful?

I also am concerned about saying that Blue Gene is the fastest computer available to humanity. Is there a problem about classified military computers, which may be faster than the computers that enter the TOP500 competition? Schoen 18:52, 17 November 2005 (UTC)


 * I agree that the Thz is suboptimal, but we are quoting from somebody else's estimate, and we can't really change that. The Teraflop number given later in the article is a common measure of supercomputer power. I'll reformulate the Blue Gene bit to address your objection. AxelBoldt 23:15, 11 December 2005 (UTC)

Server hardware and software says nothing about software
Change it to just Server hardware? New299 12:22, 17 August 2006 (UTC)

Current Hardware
Seems out of date. I'm seriously wondering if they're still using 80GB hd's on their 1.4 Mhz procs, with 256 MB ram...? 130.111.158.153 00:03, 11 February 2007 (UTC)

Server Operation
From what I know, and from the published paper on GoogleFS, I find it highly unlikely that they use RAID-1 in their production, data center nodes. The whole architecture is designed to workaround commodity hardware, using commodity hardware -- except in low capacity operations where it makes sense to use specialised hardware. 217.147.80.137 03:57, 7 May 2007 (UTC)

Article name?
Is Google platform an accepted/official term. If not, I suggest this article be more properly (syntatically) called Google's platform, or even better (from the PoV of clarity), Google's infrastructure/Infrastructure of Google. If it is an accepted term this needs to be reflected by introducing the term in the opening as is done on a normal article. &mdash;Joe Llywelyn Griffith Blakesley talk contrib 03:16, 2 July 2007 (UTC)

Inconsistnency
The Current Hardware section seems out of date with regard to the lead (in fact I am pretty sure it comes from a 2003 document). 1Z 19:28, 3 August 2007 (UTC)

Google Platform images
I noticed one of the graphics on the page which shows the load balancing does not have English text on it. Since this is the English Wikipedia, it would be nice if the original graphic designer could permit it to be translated or an English equivalent supplied. Granted, most people reading this article will know what the image means; however, it doesn't mean that it is acceptable to have non-English text when an English equivalent is more than possible. —Preceding unsigned comment added by 63.161.86.254 (talk) 19:00, 6 June 2008 (UTC)

Bizarre Supercomputer Claim
"According to this estimate, the Google server farm constitutes one of the most powerful supercomputers in the world. At 126–316 teraflops, it can perform at over one third the speed of the Blue Gene supercomputer, which is (as of 2005) the top entry in the TOP500 list of most powerful unclassified computing machines available to humanity."

What nonsense. Can anyone think of anything less mad to say on the subject? I am reminded of claims that the IBM Cell is a supercomputer on the basis of its FLOP score. Rsynnott 00:57, 3 March 2006 (UTC)

The Top500 has its own degree of nonsense; it is based purely on a Linpack score, not real-world applications. That being said, the Google server farm would probably have trouble even getting a high Linpack score because of low-performance interconnect (GigE instead of InfiniBand, etc). Google server farm is a massive web server, not a general purpose supercomputer. --Westwind273 (talk) 19:08, 23 June 2008 (UTC)

greenwashing google's Council Bluffs farm
The comment about google trying to be green by locating in Council Bluffs is unsubstantiated by the citation provided. That reference only comments about the energy resources in the area being desirable, not that they were green. Considering that the largest coal fired power plant in Iowa just opened up in Council Bluffs last year, the easy answer is that there are in Council Bluffs because of cheap and dirty coal-fired power.

On the other hand, google has made positive environmental moves as a company and through google.org. Their "Renewables cheaper than coal" effort is aggressive and worthy of recognition. But that's not relevant to this page.

The text of the page needs to have the "green" reference removed or substantiated. —Preceding unsigned comment added by 144.62.23.240 (talk) 14:38, 28 August 2008 (UTC)

Goobuntu?
Goobuntu redirects here, but there is no mention of it in the article. There is a link from at least the Ubuntu article. What is it?Otus (talk) 10:37, 28 September 2008 (UTC)

Just thought I'd echo Otus' cry: I also came here seeking Goobuntu, but it's not mentioned at all here (nor anywhere else). Warrickball (talk) 15:40, 14 January 2009 (UTC)


 * The article was redirected back in October without discussion. I have reverted the redirect and recreated the article. ~ Paul T +/C 20:14, 14 January 2009 (UTC)

Sources?
Very little on this page can be cited; each fact should be linked to a Google 'public annoucement'. e.g. has Google published how high their cooling towers are?

How could the image showing Google's first production server rack be taken in 1991 when a Canon EOS 5D was used? Just curious, but the camera was released on August 22, 2005 which proves that the metadata is false.. —Preceding unsigned comment added by 81.229.97.97 (talk) 16:18, 17 February 2009 (UTC)

Interesting Article on Google and the Linux Kernel
I don't have time to add this into the article, but maybe someone else would? http://lwn.net/Articles/357658/. Basically they run their own customized Linux kernel, starting with 2.4.18 then to 2.6.11 - primarily because they needed SATA support - followed by a 2.6.18-based kernel. They are now working on preparing 2.6.26. 74.14.15.65 (talk) 16:34, 8 November 2009 (UTC)

Modular Container Datacenter
The most glaring issue with this article being out of date is that Google has been moving to a containerized data center since 2005. I'll start listing some links for research, and taking notes here. The main article should at a bare minimum have a section on the containers.
 * Google Modular Data center - this article should link there at a minimum, and perhaps this material should be on that page?
 * April 1, 2009 - http://news.cnet.com/8301-1001_3-10209580-92.html - closeup of a Gigabyte GA-91VDP motherboard with 4GB RAM, 2 HDDs, a PS, and a UPS.
 * April 07, 2009 - http://www.youtube.com/watch?v=zRwPSFpLX8I - tour of their first data center (2005?) focusing their best practices to minimize Power usage effectiveness:
 * - cooling towers, generators, low voltage power distribution,
 * - hangar of 45 containers, 1000 servers each, consuming 10MW.
 * - As in the c-net link above, servers have a per-tray UPS
 * - air cooled (this is a surprise considering the Sun Black Box containers are water cooled) but with hot and cold air separation.
 * April 10, 2009 - http://www.theregister.co.uk/2009/04/10/google_data_center_video
 * Oct 15, 2008 - http://www.theregister.co.uk/2008/10/15/google_and_intel - container data center patent application in 2003.

--Mnp (talk) 17:17, 28 November 2010 (UTC)

New lead proposed for this article?
"Google service and product offerings have greatly extended from its original and predominant search engine []. Now Google is a major player in online advertisement [], cloud based services (Google Apps for Business [], Gmail [][], Google Docs [], Picasaweb ]), content hosting (Youtube, Google AppEngine, Google code) [ref] and is also entering the Social Networking business (Orkut at first, and now Google Plus) [ref]. But how can Google offer all of these services across the world whilst provding the best possible user experience? The answer is the platform - the common base (hardware + software + network) that Google uses to hosts its services on."

And then we could go into the article with up to date details from Google public information (for example historical data can be extracted from [] and platform data is available at []).

Ludovic.ferre (talk) 07:27, 28 September 2011 (UTC)

Original Locations?
The article states that the server farm was originally spread out over "four sites: two in Silicon Valley, California and one in Virginia." That accounts for three of the four sites. What was the fourth? The cite is a book, so I haven't checked myself. Meviin (talk) 05:04, 23 December 2011 (UTC)

1. April joke?
Hi, are you guys sure that the "information" from the 2. April are real? Up to now we had no luck finding any sources. —Preceding unsigned comment added by 80.131.181.219 (talk) 09:57, 2 April 2009 (UTC)


 * This is a reliable site, The CNEWS. Everything, published by CNET is considered Reliable Source. `a5b (talk) 07:14, 29 July 2011 (UTC)
 * The same information is here: and this book is one of most reliable sources in the field. Author of the book is board member of Google . `a5b (talk) 02:02, 14 June 2012 (UTC)

TOR?
What is the story of the "top of rack" (TOR) acronym? Some aspect of an Ethernet switch? I thought this was just a marketing buzzword, but I wonder if Google coined the term? Or was it Cisco or someone else? Any good definition? I saw it appear in the some Facebook documents too, for the Open Compute Project. Ran across this document from Cisco from 2009 but wonder if it was coined before? W Nowicki (talk) 00:02, 10 July 2013 (UTC)

Google Cloud Platform
https://cloud.google.com/security/security-design/ provides quite a bit of information on the infrastructure. Nemo 22:33, 15 January 2017 (UTC)