The Computers Powering Computable Knowledge

May 12, 2009
shadow
Wolfram|Alpha Infrastructure Team
Posted by

When Wolfram|Alpha launches, it will be one of the most computationally intensive websites on the internet. There is no way to know exactly how much traffic to expect, especially during the initial period immediately following our launch, but we’re working hard to put reasonable capacity in place. Will we have enough computing power to provide computable knowledge for everyone who visits? We hope so.

We’ll service Wolfram|Alpha from five distributed colocation facilities, which we somewhat unimaginatively call locations 0, 2, 3, 4, 5 (1 as a backup). What computing power have we gathered in these facilities for launch day? Two supercomputers, just about 10,000 processor cores, hundreds of terabytes of disks, a heck of a lot of bandwidth, and what seems like enough air conditioning for the Sahara to host a ski resort.

One of our launch partners, R Systems, created the world’s 44th largest supercomputer (per the June 2008 TOP500 list). They call it the R Smarr. We call it the most powerful computer that will be running Wolfram|Alpha on launch day!

Dell is another of our launch partners. They helped us pull together in just a few weeks a data center full of quad-board, dual-processor, quad-core Harpertown servers. That’s the computing power of another top supercomputer in its own right!

What does it all add up to? The ability to handle 175 million queries (yielding maybe a billion) per day—over 5 billion queries (encompassing around 30 billion calculations) per month. Big ambitions? We hope it’s more than enough capacity even at the busiest peak times of the day, and more importantly, that it will give you a great computable knowledge experience.

So, we hope we have built a strong, flexible infrastructure… and that you’ll be delighted when you try it.

23 Comments

This is a big time for ai running on supercomputers. Is the ibm project calles Watson somewhat similar to wolframalpha or do they use a different approach?

http://www.research.ibm.com/deepqa/

Posted by Kai May 12, 2009 at 12:30 pm

    Just a point on supercomputers. I think the Cray systems would cope with this, superscala and affordable to an institution such as this.

    Posted by Nigel May 12, 2009 at 2:33 pm

Waw !!!! I hope is enough :) !!! Nice job,guys,we appreciate !!! Can`t wait for you to “pimp” our internet experience.
I have a few questions:

1. I saw in the last time,a lot of internet sites down(for a period of time) because of denial-of-service attacks (DoS attack),how vulnerable is Wolfram Alpha to this atacks?

2. Will we be able to integrate an Wolfram Alpha search box in our personal internet sites ?

Posted by Ionut Danet May 12, 2009 at 12:36 pm

    Hi Ionut Danet,
    These are great questions. For the first question, we are preparing as best we can and are looking into having a post about this in the future. For the second, the home page will be linked to several different areas. One such area is called “Downloads.” Please check there once we are live to see if any of the options will fit your needs. Thanks!

    Posted by The PR Team May 13, 2009 at 11:38 am

Had you tried in spanish ? Relacion entre la Clorofila y la Salud Humana? or French ? : chlorophylle et humain sante ? or russian:хлорофилл и человеческое здоровье among other s? Or do you think that the only language is english ? Perhaps there is no info available in english language but that does not means there is no information at all. You should expand your search capabilities. regards.
G

Posted by Godoy May 12, 2009 at 12:39 pm

    Hi Godoy,
    Currently we are working in English, but you will be able to query different languages. Once we’re live, simply type in Spanish, French, Russian, Chinese, and so on to see interesting facts on those languages. Thanks!

    Posted by The PR Team May 13, 2009 at 11:34 am

How is the data updated/new data added and how often?

Posted by David May 12, 2009 at 12:54 pm

    Hi David,
    Good question. This is a topic we intend to do a post on soon. Stay tuned!
    Thanks!

    Posted by The PR Team May 13, 2009 at 11:35 am

I’m sure that you are ready for everyday use. The horsepower is incredible, and I suppose that you will use some kind of cache for frequent questions (you know, “meaning of life”, “existence of god”, “1+1″…)

But I would expect a total overload the first hours or days and a lot of hacking attempts, that will last forever, I’m afraid.

I really expect the best for Alpha and the Alpha team and I’m a little envious of this “one time in the life” moment you are experiencing.

Long live to Alpha!

Posted by igo May 12, 2009 at 12:57 pm

“The ability to handle 175 million queries”

that is PER DAY i suppose, otherwise this will be an expensive project ;-)

Posted by sander huisman May 12, 2009 at 12:58 pm

    Thanks! It is indeed per day.

    Posted by The PR Team May 12, 2009 at 1:54 pm

Dear People at Wolfram | Alpha,
These are exciting times. You are God’s intelligent magnanimous people truly at work here on this earth.
These phrases might be the future slogan of Wolfram | Alpha:
The more you think, the more you Wolfram;
The deeper you research, the more you love Wolfram.

Suggested Features:
Some suggestions that might be added:
Provide a trend type of graphing much much better than Google Trends, at least with numbers and counters.
Provide a kind of MDS (Multidimensional Scaling) graphs with selectable axes, to show white areas.

Please do not keep us waiting so long.

Posted by M. Mehdi Sepehri May 12, 2009 at 1:28 pm

Wow.

Fascinating and absolutely impressive . . .
This is the next step. The next century. The next period.
Keep on like this.

Posted by Rainer Kleber May 12, 2009 at 1:59 pm

Will you address real time fresh Search/Q&A as well?

I mean questions like “what is the rate of microsoft corporate bond?”

Posted by Antonio Gulli May 12, 2009 at 2:15 pm

This is a formidable idea. It’s what was missing on the web. We will judge you by the results, however. Good luck ! You deserve it, because yours is a great idea. It really could be the next step.

Posted by Bruno May 12, 2009 at 3:14 pm

Wow, 175 million queries a day. Tough to say whether that will be enough for launch day or not. Almost certainly that will be enough once the initial wave dies down, but who knows how many people will wander over to Wolfram on the 18th! I can see people running a few dozen queries, so things will add up fast.

Posted by Daniel Bigham May 12, 2009 at 4:19 pm

I guess wolfram alpha would be the first true engine computing site serving various needs across the globe.

Posted by sule May 12, 2009 at 5:12 pm

Nice post

Posted by Limousine Houston May 13, 2009 at 12:37 am

exciting, Great expectations are there

Posted by nagendra Prasad May 13, 2009 at 1:08 am

Very much looking forward to

Posted by 景鹏飞 May 13, 2009 at 2:49 am

Wolfram|Alpha Group announced that all answers are from database, which was developed many years. From my point of view, it’s not right to collect facts and then return them in form of answer. Even in a very detailed one. The real intelligent machine ought to generate answer “on the fly”, even if it does not have needed facts in database. Today it’s too very simple to know facts, which were discovered earlier. The real value is to develop a method to discover still unknown, non-described facts. That’s the point.
Such intellectual search engine is described here: “Universal Scheme of Evolution and Search Engine Forecast,” http://www.triz-journal.com/archives/2008/11/04

Posted by Alex Zakharov May 13, 2009 at 8:02 am

[...] other challenges, made particularly interesting by the sheer scale and complexity of launching two supercomputer-class clusters along with three other [...]

Posted by Wolfram|Alpha Blog : Counting Down May 16, 2009 at 8:49 am

[...] We first introduced the public Wolfram|Alpha technology on May 15, 2009, through our free, public website. Wolfram|Alpha represents hundreds of man-years of capital and effort to build, and today is maintained by hundreds of staff continually adding and verifying data and algorithms, and is backed by a supercomputer-class infrastructure. [...]