Slashdot Mirror
Google Moves From Search To Inventor
TubHarsh writes "The New York Times reports that Google continues to expand its scope from search engine to inventor. Google assembles the majority of the hardware it uses and deploys at such a large scale, that Google may be 'the world's fourth-largest maker of computer servers, after Dell, Hewlett-Packard and I.B.M.'. The article also states that Google may be entering the chip design market with new employees who were ex-Alpha Chip engineers."
Now to comment on something I read in the article: I disagree with that. I think it should be re-stated to say "It is very difficult to accomplish more than you have the resources to sustain." It's fatal in thinking that you only do one thing for a business to be successful. A simple analogy would be the farms that I grew up on. No one specialized in one crop or animal. Why? Because sometimes the market would tank for one particular thing and it would tank hard. If you had a distributed investment in produce (like a portfolio) then you would survive most of the market problems. I think Google's strategy is much the same in that they are trying to cement themselves in other technologies--not because they're going to lose the search market--just because it's a smart thing to do.
I think that there's a lot to be said about concentrating on one thing and getting it right. If you do get it right, then it's encouraged to move on to something else. I think Google has found themselves in the top of the search engine market. They found out that their technology doesn't work so well for closed domains (military or business level searching) so I think they just need to keep looking for new ways to stay ahead of the competition. Meanwhile, they have seemingly unlimited resources. Why not try to build your own router?
I mean, fresh graduates are cheap. Some fresh graduates have a lot of ideas and are decent workers while the majority of others are lemons that don't do anything. Why not hire a bunch of them and spend a lot of money weeding them out? I think it's great that Google's taking a stab at other technologies and I honestly think they have a good strategy for doing it.
To comment further on the article, Google makes unreliable machines reliable en masse via redundancy. They are indeed very secretive about their technology but if you want to learn more about their page ranking algorithms or basic technologies, why not read their patents? They always seem to be covered on Slashdot anyway.
My work here is dung.
Now that does interest me. If they can show the same level of industrious innovation that they have in other fields, I'm excited about the impact this may have on the server-market, if nothing else.
I just hope that, if they are developing chips in-house (and if they are, I expect them to be cheap and powerful), they are less tight-fisted than they are with their other technical innovations. A new power-player in the CPU market would be great for us end-users
Seriously though, if they start manufacturing all their own hardware from scratch, they're probably going to be more independent than any major computer-based international in recent history. *exaggeration ends*
Meta will eat itself
Much as I'd like to see the Alpha return, backed by Google (or pretty much anyone else. The death of PALCode was a sad day for the industry), it doesn't seem likely. The Alpha approach was to build the fastest chip possible; in terms of performance-per-watt or performance-per-dollar, it didn't do so well.
I am TheRaven on Soylent News
Friends, remember that Google is the America hating empire.
This new wave of innovation probably uses Linux (created by a European communist) with a sordid history. No doubt this is part of an insiduous plot to destroy the valuable patents of The Sco Group.
Their so-called "inventions" have already led to a huge upturn in hacking, eponymously named "Google Hacking". All true patriots must support tougher sentences for such evil terrorists.
...welcome our new Google - Cyberdyne Systems overlords.
When will the Terminator-1 chip have been designed ?
I don't think we should think of this as a move that Google may 'sell' the machines they make, aside from selling Google search or app appliances one day. The vast majority of chips they would be making are probably to 'own the supply chain' for their own massive server systems. This is similar in concept to the early Ford Motor Company that owned the steel mills, etc. Google just wants the lowest net cost per computing cycle, and if Dell wants to earn a profit selling them computers in bulk, it might be cheaper for Google to bring that profit in-house.
stuff |
I think they're just throwing stuff at the wall and seeing what sticks, much like VCs do. But their doing it all in-house, hoping to come up with the next big thing. And the thing after that.
It's not offtopic, dumbass. It's orthogonal.
It was part of a university to begin with.
perpetually dwelling in the -1 pits
http://news.bbc.co.uk/2/hi/technology/5140066.stm go figure
Google is an innovator in many ways and I believe this may scare the most unoriginal Software company ever Microsoft, we can only hope there is a Google OS coming soon!
TheADDkid.com
I think google is just an amazing company, they hire some of the worlds top developers, build their own servers and apparently their own cpus now just to make sure that everything runs smoothly. Couple that with froogle, google maps and google earth, summer of code and submission of code back to open source projects such as wine. It's a shame that there aren't more companies like google that do everything they can to put their customers first and their profits later.
They are indeed very secretive about their technology
Yes, so I thought. And indeed the article says, "Google is notoriously secretive about its technology", "Google will not comment on its costs". Yet Bill Gates is quoted as saying "Google doesn't have anything magic here. We spend a little bit more per machine. But to do the same tasks, we have less machines.".
A web search doesn't turn up the reference for that quote (and the article doesn't link to it), so it's hard to know the context. But still, it does seem odd. How can Gates know such details, which are supposedly secret? I don't know whether to doubt the truth of his claim, or to wonder about how he could have found it out.
As Google grows and start giant projects (like Google Earth) it requires more and more resources. It is natural if they start building own computers, it will be the base for new projects. I like Google because they are not afraid of global large-scale projects.
Hide your files and folders from others!
They also probably reduce thebrain-drain of their talented employees - since working on Google must be very, very rewarding for someone with an imaginative mind but not a lot of organizational know-how.
Stop the brainwash
Cliff Claven
K.E.G. Party Chairman
Founding Leader of: Koncerned for Egalitarin Governance
IBM had server sales of more than five billion dollars last year (or three billion, if you don't count mainframes). Even lowly Sun beats out Dell, which comes in at almost $1B.
Keep in mind that this is just for one year. Pick your favorite guess for how large Googles server farm is and divide by the average age of those machines. Do you still think they're assembling more than a billion dollars of hardware per year?
Am I part of the core demographic for Swedish Fish?
Can Google become an artificial intelligence?
Google certainly has the data to whet the appetite of an AI Mind, but first Google would need an AI Engine such as Mind.Forth to impose order on the data, so that Google would not just store the data but would know the web of data.
Maybe Google will trigger a Technological Singularity.
Hardware is where DRM resides and operates. Why would Google buy DRM from HP, IBM or Dell? With all that talent at their disposal, Google would design their own brand of DRM... hell, even the chips and architecture... weeding out useless buses, registers and stuff.
We already know they use a modified Linux kernel... what better than using proprietary hardware as well? That way, they are free from the cluctches of Intel, AMD, ATI, NVidia, HP, IBM etc., besides Microsoft, Oracle and the software gorillas.
If you keep throwing chairs, one day you'll break windows....
But then just how many specialized chips does one need? With personal computers it's getting a little out of hand. First we have graphics processors, and now physics processors. Oh, and we also have network cards that allow you to offload the entire TCP/IP stack to their own processors. Oh, and sound cards have hardware mixers, so you don't have to mix the sound in software on your general purpose CPU. Oh, and those video capture cards convert everything to mpeg in hardware, so you don't need your CPU for that either. All I need is a special processor for compiling code, and I could go right back to using a 486 as my main processor, since it wouldn't have anything to do anymore.
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
Is it just me, or does the microprocessor business seem a REALLY bad one to get into right now. Maybe 30, 20 or even 10 years ago, it wouldn't have been such a bad option. But with Intel and AMD going forward in this perpetual juggernaught race, it seems like anyone getting into this business is Dead On Arrival. Transmeta Corp anyone? Dan
- There's no place like 127.0.0.1
The Google Alpha Beta
You say that like it's a bad thing...
Everyone is foolish in only seeing the search aspect of Google. The reason search is so prevalent on our minds is that Google took advantage of the greatest tech out of the gate, unlike MS, and made it the best first. Now they have time to focus on everything. We have to reevaluate our views on technology from desktop centric to the Network like Sun says. Google made sure search was done right first as it is the most important technology Google is the phone company! It's great they are creating chips with Sun hopefully open source as I would think they are using the Sun Spark program. The future looks bright.
On several occasions I've suggested to customers that they consider building their own servers. Going by the look on their face you'd think I'd just asked directions to Mars. I'll usually let them ba-humbug the idea for a while before informing them that Google does it and always has. That usually gets them started asking questions instead of telling me why it's such a bad idea.
That wouldn't work for most companies, but if they've got a technology core group that's big enough or if they're a tech oriented company, it's the most flexible way to go and not that much more work. Some standard box configurations and parts lists keeps all the components working together.
I wouldn't even consider buying a server for my own stuff. If there were an ATX type standard for laptops, I'd build those as well.
That's our life, the big wheel of shit. - The Fat Man, Blue Tango Salvage
Yes, it is a bad thing, because a regular processor only costs $150. On the other hand a video card and a physics card and all the other cards you would need to run a computer this way would cost much more money.
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
For Google, the sums are completely different. Every watt Google uses on a CPU costs them twice; once to turn it from electricity into heat (via computation) and another to extract and dissipate the hear. In California alone, the amount of power to be used by proposed data centres (i.e. those to be built this year) is over 10% of the power consumption of the entire state.
Any CPU owned by Google is likely to be running at 100% capacity a significant amount of the time. If you can use 10 custom chips instead of 100 Opterons, then the cost savings are going to be huge. Also, remember that these are in-house chip designs that are being proposed. The cost of making one chip is huge. The cost of making ten chips is exactly the same as the cost of making one. The cost of making one hundred is only very slightly more. If Google need a few tens of thousands, the numbers start looking a lot better. If they can sell a few hundred thousand more (at a 50% markup, of course), then this is even better; their costs go down, and their competitors are paying more for exactly the same thing.
Power management in datacentres is a huge research area; the International Conference on Autonomic Computing this year had an entire section dedicated to this field.
I am TheRaven on Soylent News
If Google's need for server is as big as it appears, it makes sense to design and build custom servers for themselves. But my gess is that they'll be following the Sun Niagara architeture, with multi-core/multi-processor designs.
;-)
My point, Google doesn't need raw performance, they need an architeture that scales well, supports lots of concurrent requests, and consumes very little power to make cluster mantaining costs less expensive. Well, thinking this way, at this point it looks like Google might buy Sun to get their Niagara processor
In the end, I don't think such a processor would be seen on a desktop computer.
---- You know how some doctors have the Messiah complex - they need to save the world? You've got the "Rubik's" complex
First of all, I was talking about home computers, so my point still stands, but to counter your argument, here it goes. The processors they would be building wouldn't be on the same market as Opterons, Xeons, Althons, or Pentions. Unless they have the same instruction set as the others, or can be programmed in the same way, then I doubt that there would be a lot of people who would want to buy them. And if they were general purpose enough that they could be used then they probably wouldn't be that good for Google anyway. Maybe MS could make some use of these processors, but I don't think they would buy from Google. Or Maybe IBM. Wait they can make their own chips. I doubt that Google could make any money selling specialized processors that are fine tuned for their own needs. None of the console makers even make their own chips anymore, and they sell way more consoles then Google could ever hope to have servers in their datacentres.
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
WTF is "Driad", Gates claims it's Microsoft's answer to MapReduce?
Also, sadly the article does not mention that Google runs almost entirely on Linux. There's room for a couple of Bill Gates quotes on how Microsoft's solutions are better, but no mention of the fact that Google has no need for any of them.
it's a blue bright blue Saturday hey hey
What is Google doing with Sun?
Internet based word processing and spreadsheets, email on the internet, a google service for everything else... It wouldn't surprise me if the next generation of personal computers are nothing more than a SunRay type thin client plugged into the internet, Sun helps with the hardware and google services will do the rest... it seems to be the vision of both companies...
Doesn't the NYT's article read just like a re-hashed Google press release with a sprinkle of 'bad' quotes about Google just to appear journalistic?
I expect better journalism from the NYT.
I was going to jump in with the same point. Google's hired people from DEC and their primary concern in their data warehouses is undoubtedly performance and energy efficiency. It's very likely any innovation in either area that would be dramatically superior to the best offerings from Intel and AMD would not be x86-compatible.
Unless the chip was so outrageously fast that they could run an x86 emulator in it without a performance loss, it would be dead in the home and office PC market.
Or Google wants to release its own line of PCs with its own CPUs and their own Google Operating System. Not very likely.
Hardly. The trend has been fewer dedicated processors in recent years, not more. Sound cards rarely have hardware mixing anymore except at the high end, instead the operating system is expected to do it. Physics processors are too early to call but a bunch of people seem to think they aren't necessary given that video cards can do a lot of similar work. Modems lost a lot of their circuitry to software. Etc. The only real dedicated chip that came into its own lately is the GPU, and that's because it has a very different general structure to a CPU.
One thing I noticed that no one even thought of as a reason is the biggest single expense google has. It's called Electricity and if they can design a cpu for their file servers that is as efficient as the Via Epia/Nemiha chips 2watts, they'd save lots of money that can then go towards other needs.
Now I'll make a blind guess as to what percentage of Googles income goes towards electricity: At least 25% of the income Google makes goes towards electricty consumption. Now if you estimate that 80% of googles consumption id related to their servers, what happens if you reduce that demand by 50-75-80%? Now you're talking a serious release of cash flow that can be used for what ever they want.
Now the $0.23 cents question: Is it worthwhile for google to design their own chip? Nothing says they have to manufacture it, just design it and I'd bet AMD/Intel/Motorola/IBM/Samsun/Toshiba and any other chip manufacturer could fabricate the damn things. So they'b be idiots to build their own fabrication plant, just contract out to have it manufactured to someone who's got the facilities and have an NDA with ironbound contract to back it up.
Cool! I'm an inventor! Yeehaw!
I knew it! Of course, we all did. I knew Google was going to get into computer sales sooner or later. I wonder when they're coming out with their Linux distro... you know it's coming.
It's rather interesting to see all the economic geniuses get together here on Slashdot and tell the world just how good their plans are. Integrate this, expand or die, etc. Is this new thinking? No. Has it been tried before? Yes. Most successfull as a whole is perhaps the Japanese Keiratsus (or how ever you spell that). Conglomerates in other words. How many of those do you see these days in the western world? How many are profitable? GE could be held as an example of success until lately. Don't quote me on it, but I seem to remember them strugling with losses as well now. What about the Japanes then? Well, they are slowly getting around to it as well. Not by choice I'd say, but dragged to the conclution by economic reality.
If you read nothing else in this post, at least read this.
The thing that most people don't get is the fact that you can integrate vertically until you are blue in the face, you still have the exact (more or less) same demands for ROC. The logical error goes something like this: If google owns the CPU manufactoring part as well then they can buy the CPUs at cost. NOT TRUE. I'll say that again in case anyone missed it. NOT TRUE. You see, the investors, whos money google would use to build this business, will have the same demands on interest as the investors that let AMD play with their money. You can pretend that you can buy "at cost" but it is never possible. In the end it always catches up with you.
Now you can stop reading.
Sure, there are economies of scale and such. Those are much better put to use by integrating horizontally though. And that has a nasty tendency to hurt consumers down the road.
The same goes for "expand to new areas or die". Sure, that is true. The thing is though, do we want a company to survive if they are not profitable? Why not have two companies that each do A and B respectively instead of one that does both A and B. As can be seen with Microsoft (since it's a good topic to bring up at Slashdot) there is little point in having one area subcidizing another area. Let the investors put half their money in company A and half in company B instead of having it all in the same company. Worst case scenario is that you do AS WELL as a combined company. Again, that is worst case scenario. There can only be better. Stop getting so attached to company names, because that is all they are. Names! The people in them come and go, the name remains. Who here has started to hate certain parts of Google because they now have MS execs? I'd bet none. Why? No reason!
Karma: 2.71828182846 (Mostly due to small, fun pills)
Maybe one day we have a GPU (Google Processing Uni) inside our PCs that has special hardware support for indexing, retrieval and text processing in general. Independently of Google or any particular vendor, the theoretical question that intrigues me is: what operations would you like to have built in to aid the search business?
PageRank in microcode? Porter stemmer as an assembler instruction?
For several decades, CPU design has been driven mostly by traditional numerical concerns. While ranking algorithms certainly are based on numerical principles as well, it remains to be investigated whether there are operations that are worth providing at hardware level, or (more likely) completely new architectures.
Note that their MapReduce paradigm of parallel data processing is close to data flow machines in some sense, and while these were not a success at the time, times have changed (it's always a question of boundary conditions).
Okay, you agree that Google is a good company, but now you are going to tell us some ways in which Google is bad, so that overall Google is not as good as we may think. I'm listening.
So, they profit a lot from ads. I was expecting something bad. Is there something inherently bad about ads? I find ads in general to be obtrusive and annoying, but Google's ads are not. Is it more an ideological stance against ads? (I'm assuming that the thing that's bad is the ads, not the making a ton of money.)
That comparison is so off-base! You're comparing apples to religious denominations. "Closed source" and "open source" would apply if they were distributing the software (as in "Microsoft Office and Windows"), but Google isn't distributing their search engine. You're using the negative connotations of "closed source" to create a misleading impression.
You might as well be saying, "John hasn't worked a single day in his life, pays no taxes, and yet lives in daily comfort getting thousands of dollars in government spending." "Really? Who's John?" "My twelve-year-old cousin."
(And don't call me baby cakes.)
Wrong.
See? I, too, can make one-word dogmatic pronouncements. The morality of Google's actions in China are hotly debated here on Slashdot and elsewhere. Since you don't bother to qualify your one-word non-sentence, I won't qualify mine.
Here again, you imply, without explanation something inherently bad about the situation. So what if we are not their customers?
I think of Google as a broker: they profit by connecting us to what we need. Just as a real estate agent connects you to the sellers of your dream home (hey, you the buyer are not the real estate agent's customer!) or a headhunter connects you to that job you've been looking for, so Google connects us to the information we need. This is one connection, but they don't directly profit from it. The other connection is that they connect their advertisers to us. They do profit from this.
It sounds like you are opposed to them "harvest[ing] us and sell[ing] us to their advertisers", but Google only profits when you follow their offered links. You have a choice about whether to use Google as a search engine, and even when you do, you have a choice about whether to follow their offered links.
It's like you're saying, "Damn, I asked the real estate agent to find me a decent house, and she did! I mean, those agents just harvest us house buyers and sell them en masse to the sellers! Damn those real estate agents for doing what we ask!"
I agree that Google is not perfect, but nowhere in your posting do you say why you think so.
404555974007725459910684486621289147856453481154 in hex is "You sank my Battleship?"
[GPG key in journal]
Why? What advantage did it offer over doing the same functions in low-level OS software, for example?
Netezza has already been doing exactly this in their product line. They use FPGAs on each compute blade that handle disk access but also handle database record layout, and can do where clauses and projection while streaming from disk.
See: http://en.wikipedia.org/wiki/Netezza
I also believe that intelligent disk IO will be important in the future. No, it's important right now. Disks are a bottleneck and having dedicated hardware that can leverage spindle parallelism and free up the memory disk controller pathway is very important.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
IIRC it's all Northup Grummond now.
And they in turn outsource the production of their computing resources to other vendors (Cray, Sun, Supermicro, etc.)
I think it's a bit of wising up on their part. Why should they be so suspicious of what are now anonymous, commodity products? Just get it while the getting's good, you know? Better to stay simple and low profile than high profile and complicated that just screams "Government Purhcase! Government Purchase!"
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
... that talks to your secondary storage.
A few PowerPC processors, and some FPGAs. Shouldn't cost much more than a typical hardware RAID SATA controller.
It could run a virtual FS in microcode on partitions on the disk (in addition to giving you standard RAID access to the deices). You use a slightly abstracted API to talk to it. Just throw your files into it, and it versions them, extracts text strings, etc. all on its' own. Then you can do fast search and retrieval of said content.
Man that would be cool. Google in your desktop. Google in your file servers.
Well, I wouldn't call it "Google". But like, GoogleVault. Or some kinda cool brandname like that. I'd be totally into it.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
I apologize for going off on a tangent, but does anyone know if there's a meaning behind the colors of the letters in Google's logo? They don't follow the Roy G. Biv rule... The closest thing I could find to an answer was:d ex.html#107307127760339576
http://blog.outer-court.com/archive/2004_01_02_in
but that's not an answer.
-Rich
When is someone going to hack... um I mean "borrow" the google network of computers and run distributed.net's client on it... I wonder what kind of key rate you could get with all those computers?
True, but you're looking at two different markets. A microprocessor is a jack of all trades, master of none. And with architecture and process advances, we've been able negate some of the negatives. However, customers who need high performance would want separate chips, as dedicated hardware will always perform better than an integrated solution. Customers who only require a low price want the integrated solution.
Paper about Google Cluster, not new, but have the same kind of information of the NYT article.p df
http://labs.google.com/papers/googlecluster-ieee.
Is google automatically copywriting stuff its robots find on the net with ai now?
Its funny seeing code you wrote for a client without a contract, that belongs to you, appear on google with a copyright on it so I guess they've been doing this for a while. (my own fault for developing front-end stuff)
-John
They state that the original computers used cork, but that proved unreliable. In fact, the cork that the motherboards were on worked just fine. The problem was that the fire inspector didn't like it, so they had to switch to plastic standoffs or something like that.
I've been saying for a long time that Google should get some engineers to design things like power supplies that could obviously be done more efficiently. For a long time Google has been using custom parts (motherboards, RAM), and even buying up cheap discarded RAM chips and making their own memory modules. It makes sense that they would hire silicon engineers. They may not be planning on making their own chips, but they may be doing their own component designs.
dom
I can already see that dreadfull clippy 'Hi, I see your values have lowered' with a box at the right:
Increase Your Numbers
Increase those spreadsheet numbers now,
get an outcome with up to 500 units more!
www.wearethespreadsheetsniffers.com
--- I am known for the ones who want to find me on the net. Is that a privacy risk or a privilege? One might wonder..
By the time they are out Beta they will be called Gamma.
--- I am known for the ones who want to find me on the net. Is that a privacy risk or a privilege? One might wonder..
Wow, I've never seen so many sycophantic, hypocritical, fanboys in my life.
(The sycophancy is obvious, the hypocrisy is that Google is a *closed source* company, which is normally cause for villification in these parts.)
It's going to be funny seeing this overrated company try to make a go of it making their own cpus, os, ai chips. Sell your stock now! (Not that you guys have any, slashdotters are too poor to own any stock, working for free and all.)