Forum Moderators: open
Turn it on at [labs.google.com...]
(You might want to opt for the conservative mode tho')
Glad to see this becoming more widespread. Much bigger fan of this type of computing project that the decryption (unless it's X-box) and Seti@Home projects.
Google Compute is a small start, here's what is going to happen next:
1) Rebirth of the Free ISP - free internet access if you "check in" your idle processing power
2) New release of Linux or Windows that can take advantage of this "processing power" network so that you can enter your credit card number when you want to run a virtual "dual processor Pentium 8" or something like that
(if some Linux company gets the patent on this before Microsoft does, Linux will unseat Microsoft as the majority OS in 3-5 years (if Microsoft doesn't buy them out)
3)Intel patents and releases new distributed computing Pentium chip that takes advantage of this processing power network at the processor level, making it even faster & more efficient. Intel now no longer has to spend $$ on R&D to keep up with Moore's law.
4) Or if Google is a step ahead, they patent the technology and come out with their own operating system and GoogleOS unseats Microsoft & Linux and Google offers $10 PC's through Wal-Mart running on a Pentium I processor but outperforming Intel's fastest chipset if you use GoogleISP.
If someone actually ever succeeds in building the Food-aracacycle it will be Google.
1) Rebirth of the Free ISP - free internet access if you "check in" your idle processing power2) New release of Linux or Windows that can take advantage of this "processing power" network so that you can enter your credit card number when you want to run a virtual "dual processor Pentium 8" or something like that
(if some Linux company gets the patent on this before Microsoft does, Linux will unseat Microsoft as the majority OS in 3-5 years (if Microsoft doesn't buy them out)3)Intel patents and releases new distributed computing Pentium chip that takes advantage of this processing power network at the processor level, making it even faster & more efficient. Intel now no longer has to spend $$ on R&D to keep up with Moore's law.
4) Or if Google is a step ahead, they patent the technology and come out with their own operating system and GoogleOS unseats Microsoft & Linux and Google offers $10 PC's through Wal-Mart running on a Pentium I processor but outperforming Intel's fastest chipset if you use GoogleISP.
Makes me want to say "Let's install Google on the computer - All the computers use Google now, as a result of the OS wars"
Kind of what like Sandra Bullock said in "Demolition Man" about Taco-Bell.
i'd rather do the SETI thing.
An open source search engine is the need of the hour(possible, with the crawler using distributed power of PCs around the world). The search engine has become one of the pillars of the internet and with nearly 90% of the power resting with Google, it presents another Microsoft like situation.
Open Source: is anyone listening....how about incorporating a similar feature into Mozilla for starters
i'd rather do the SETI thing.
- it give's you a pretty graph/screensaver and all :)
rustyzipper - Fascinating.. So you think we're going back to the terminal/mainframe way of computing.. just using 'all-the-computers-in-the-world' as the mainframe..
I haven't seen a great deal of the Intel X-Scale's yet, but I've some experience with the ARM chips their based on... given the enourmous postage-stamp(arm core) to large-brick(pentium) difference I recon we're gonna see a massive shift in the next couple of years to really quite powerful and cute devices... . .
Maybe in a couple of years your fridge will be spending it's day computing the meaning of life!
42 anyone ;)
2) New release of Linux or Windows that can take advantage of this "processing power" network so that you can enter your credit card number when you want to run a virtual "dual processor Pentium 8" or something like that
(if some Linux company gets the patent on this before Microsoft does, Linux will unseat Microsoft as the majority OS in 3-5 years (if Microsoft doesn't buy them out)3)Intel patents and releases new distributed computing Pentium chip that takes advantage of this processing power network at the processor level, making it even faster & more efficient. Intel now no longer has to spend $$ on R&D to keep up with Moore's law.
4) Or if Google is a step ahead, they patent the technology and come out with their own operating system and GoogleOS unseats Microsoft & Linux and Google offers $10 PC's through Wal-Mart running on a Pentium I processor but outperforming Intel's fastest chipset if you use GoogleISP.
Can't see any of these happening.
The massive network latency when compared with the internal of a computer even on the fastest possible internet connection makes distributed computing for everyday applications completely impractical. Distributed computing is great for CPU intensive tasks that can be easily split into batches and then recombined.
99% of applications are not like this.
Maybe in a couple of years your fridge will be spending it's day computing the meaning of life!
huh, it better not slack off cooling my beer while it's doing it ;)
I met the person in charge of Toolbar projects (Wesley) when the Google team dropped by Harvard a few weeks ago. I made a suggestion regarding distributed bookmark searches- a sort of incidental path weighting somewhat akin to Amazon's "people who bought this also bought", but more clustery. I hope they're considering playing along those lines.
Id like to know what language 'Xhosa' is and where it origionates ;D
But i was just thinking sometimes we can't make it too strange, less powerful search engines are standing now.
It is terrible if someday all the search engine turns the same results----even if it is highly qualified.
Thousands of computers across the globe crawling web sites randomly and sporadicly with varying IP addresses could lead to more hassle then it's worth. The constant question of "Who is this crawling my site" would become somewhat worrisome due to the potential abuse.
Google Compute lets people contribute their idle CPU cycles to scientific research, but it has nothing to do with crawling the web. Crawling the web isn't the hardest part of making a good search engine--it's what you do afterwards that matters the most. :)
You're right. But, Google could benefit from setting up a Grub like system where crawls are performed for desktops of Google enthusiasts.
The computing power and bandwidth that Google could save would be immense.
In fact, you have to assume that User A isn't trustworthy--User A might claim that all the pages he fetched have links to user-a.com, for example. So now you might also have to ask User B to fetch the same pages in order to check on User A. That also means more bandwidth costs for everyone. It's easier just to fetch the page directly. Distributed computing has some interesting qualities, but for these and a few other reasons I'm skeptical of distributed crawling.
It'd be great if Google shares the exact same opinion as you. For I had the same fears (Grub actually gives its users "preferred" rankings).
What would be nice is if Google could incorporate multiple, selectable distributed computing projects into the Google toolbar as opposed to just F@H.
It's not really a bandwidth saving
You need distributed computing becuase you know that your cost of just assembling data is super high: new, removed and fresh pages. This is not indexing...indexing comes next, first you need to know about the page.
You have a unique code for each page(fingerprint) and you know that page has changed when the code changes. You then send the crawler out. Right now you aren't able to do this successfully, your detective and your crawler are combined in one...but with distributed computing this would change the game, wouldn't it? The local crawler would be used to report the code, and if the code is different it would trigger Googlebot
You know you don't have unique algo, you just have more computers than anyone else (50,000 and counting)...but even those can't keep up with this crazy web...that's why you need distributed computing...
I don't think Google needs distributed network to crawl the web. Judging by their algorythm though, they may use some of that distributed CPU power to calculate PRs.
