Thread: Google's had quite the week

  1. #16
    Registered User
    Join Date
    Sep 2004
    Location
    California
    Posts
    3,268
    Quote Originally Posted by Mad_guy View Post
    In terms of contributions to e.g. Linux, as a company they are actually one of the smallest contributers there is, although they are built almost entirely on Linux from what I understand.
    This is false. Google contributed 40936 lines of code to the 2.6.32 kernel release. There are only 6 companies that contributed more. Source.
    bit∙hub [bit-huhb] n. A source and destination for information.

  2. #17
    (?<!re)tired Mario F.'s Avatar
    Join Date
    May 2006
    Location
    Ireland
    Posts
    8,446
    It's doesn't seem like you, taking the part for the whole, bithub.
    Neither failing to fully read what you quote:

    There are a couple of other interesting entries here. Google takes a lot of grief for not contributing back, but that company was the source of a fair amount of code going into 2.6.32. Much of that was support for the HTC "Dream" (aka G1 or ADP1) phone platform, but Google also contributed to control groups, ext4, memory management, IPVS, and libata.
    Google is notorious for not contributing back. It just so happens they did it on 2.6.32. Even though most of that was to offer support for their own business interests.
    If they keep it up from now on, then it indeed becomes false. But for now it isn't.

    So, just to annoy you a little
    Google is one of the worst contributors even though most of their business is linux based. 2.6.32 didn't change that.
    Last edited by Mario F.; 12-16-2009 at 05:31 PM.
    Originally Posted by brewbuck:
    Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.

  3. #18
    Disrupting the universe Mad_guy's Avatar
    Join Date
    Jun 2005
    Posts
    258
    Cycle Gap: Linux Kernel Development Stats from Greg Kroah Hartman

    "Google is at number 13 with 1.4% contribution. Without Andrew Morton's contributions Google's would be at the fortieth spot."
    operating systems: mac os 10.6, debian 5.0, windows 7
    editor: back to emacs because it's more awesomer!!
    version control: git

    website: http://0xff.ath.cx/~as/

  4. #19
    Reverse Engineer maxorator's Avatar
    Join Date
    Aug 2005
    Location
    Estonia
    Posts
    2,318
    Quote Originally Posted by Mad_guy View Post
    It's unfair to ask a hand-wavy question like that without actual numbers to support a reasonable claim: how much does 1 of those 100,000 servers cost? Let's make it more reasonable: 10,000. If they can get reasonably cheap commodity hardware, and factoring in that you will typically get a deal on buying in the large, let's say that 1 server out of those 10,000 costs $1,000. Then they are spending ~$100,000 (i.e. nothing) on this new hardware. Even if you bring it back up to 100,000, that's only $1mil USD, which is, to google, something along the lines of pennies in your pocket.
    1000 x 100000 = 1000000 ? Are we both talking about base-10 numbers?
    Quote Originally Posted by Mad_guy View Post
    Let's say their existing cluster is 100,000 computers: if they double that, and get double the speed increase with that extra 100,000 computers, they have effectively doubled the performance of their application by spending very little money and time, compared to the amount of time and money it not only takes for a developer to rewrite the code, but also test it and make sure it performs well and even better.
    Optimizing the code costs more than $100 000 000 (1000 x 100 000) + $?00 000 000 (the cost of building a huge server park and setting it up)? Sorry, I just don't buy that. Oh wait, I just did a small search and it seems 25% of Google's servers is about 250 000 servers. So... the total costs of adding 25% of servers and building the server parks and powering them and all other costs included, I think the whole thing would cost around half a billion dollars, if not more.
    Quote Originally Posted by Mad_guy View Post
    Here's another question: how big is the software that could potentially be rewritten, in question? If it's 5,000,000 lines, well, you're probably better off buying hardware and trying to scale upwards if your design lets you.
    Optimizing means rewriting bottlenecks.
    Quote Originally Posted by Mad_guy View Post
    Point being: there are tons of variables here you aren't considering. The question of whether or not it's "better to rewrite the software" or "better to throw more hardware at it" is not as clear-cut and dried as people would like to make it out to be.
    Yes, there are tons of variables I am not considering. But I think the variables you are missing about adding physical hardware (in quantities over a hundred thousand pieces) are more relevant.
    "The Internet treats censorship as damage and routes around it." - John Gilmore

  5. #20
    (?<!re)tired Mario F.'s Avatar
    Join Date
    May 2006
    Location
    Ireland
    Posts
    8,446
    Great, take a 1 in a Million example and use that as the Rule that establishes the truth for all cases.

    If you have any doubts, the quote is still here and will stay because
    a) it is true for most cases,
    b) It can be years before I bother changing my sig, and
    c) I don't give a f... about your 1 million servers company example. It only proves you think small.
    Originally Posted by brewbuck:
    Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.

  6. #21
    Registered User jeffcobb's Avatar
    Join Date
    Dec 2009
    Location
    Henderson, NV
    Posts
    875
    I respect Google for what they are doing with Android. I recently had a chance to work with Android at a systems level and while I am not totally in love with how they shut out all but the Java coders, I do like having a completely free software stack on a phone that you can build from source if the urge hits you.

    That said, I too hope they don't "get too big for their britches"....
    C/C++ Environment: GNU CC/Emacs
    Make system: CMake
    Debuggers: Valgrind/GDB

Popular pages Recent additions subscribe to a feed

Similar Threads

  1. Determining the Date of the [nth] [Day] in [Month]
    By Dave_Sinkula in forum C Programming
    Replies: 5
    Last Post: 05-25-2007, 04:12 PM
  2. structural fault any ideas
    By kaijuu in forum C Programming
    Replies: 17
    Last Post: 04-17-2007, 02:43 PM
  3. day of week
    By s_ny33 in forum C Programming
    Replies: 18
    Last Post: 11-02-2005, 11:56 AM
  4. Force end
    By Redattack34 in forum C++ Programming
    Replies: 9
    Last Post: 09-06-2005, 11:16 PM
  5. The day of the week America was discovered
    By Gustaff in forum C Programming
    Replies: 3
    Last Post: 08-12-2005, 09:47 AM