Can the mother of all supercomputers save us from Big Brother?



Today I’m annoyed at Facebook. Among the amazingly witty and touching postings from my friends and Amnesty International are “pages you might like” and advertisements for things I don’t need, especially single women.

But on a happier note, I’m thinking about the mother of all supercomputers (and if there’s one thing I’ll always be grateful to Saddam Hussein for, it was for introducing into the Western lexicon the expression, “the mother of all …”).

As I speak, the internet is enabling me to be simultaneously sitting in on an international advisory board meeting from the privacy of my own bedroom using a webcam, running jobs to look for neutron stars on 57,344 processing cores on my new toy, Swinburne’s “Gstar Supercomputer” (or what I like to call “the beast”) and of course, thinking about this article.

I learnt long ago that when you rock up at a university and tell people you are an astronomer it sounds a bit esoteric and irrelevant, but when you say you’re a supercomputer expert it gains traction, funding and subsequent glory!

Why? Well fortunately, most senior bureaucrats in universities like to invest in enabling infrastructure, and supercomputers sound like serious, hard-core investments that will lead to engineering and scientific glory.

Supercomputers can also be turned off at no cost, and because of the wonders of accrual accounting their impact on the budget is minor.

Enter the beast

At Swinburne we have an astonishingly-powerful supercomputer. It has more than 100,000 cores, a 2,000-terabyte “disk” and 10,000GB of RAM all linked by 40 gigabit-per-second networking.

It is the sort of machine that brings a tear to an astrophysicist’s eye. We can literally simulate the universe in it, make movies, and find neutron stars in astonishingly peculiar environments.

The beast’s architecture and “building blocks” are not unlike the machines used by Google and Facebook in large data centres. It comprises about 150 “enterprise-grade” GPU servers on an uninterruptable power supply in a custom machine room.

Over the next few years it will grind through Petabytes (1PB = 1024TB) of data and serve the Swinburne and wider Australian astrophysical communities, courtesy of Astronomy Australia Limited.

But the beast is but a mouse compared to the cumulative domestic computational power plugged into the internet and owned by Mr and Ms Average in their own homes.

And there are some computational problems, such as searching for aliens, that are “embarrassingly parallel” – that is, a problem divisible into so many independent parts that can all be run in parallel that I’m embarrassed to say how many – a fact recognised by the geniuses behind SETI@home and BOINC.

The SETI@home project has used more than 3 million home computers to look for aliens, and is the largest distributed computing project ever devised. A computational marvel. A triumph of software engineering.

And the current alien body count? Nil. Even in Roswell. But even SETI@home is only using a tiny fraction of the “mother of all supercomputers”, the global internet and every home PC on the planet.

Facebook is watching

So what other problems apart from climate research, cosmology, gravity and alien-hunting map to this massively distributed paradigm? Well, maybe Facebook?

Facebook houses a repository of everyone’s friends, postings, photos etc. When you log in, Facebook builds a web page that lists your friends\’ posts, their photos, and responses to your friends\’ posts. It is an embarrassingly parallel problem with a bit of data exchange.

It also works out from all your data what advertising you are most likely to respond to and hits you with ads. It knows who your friends are, and what you like. It spies on you to know what to try and sell you.

It tracks you around the world. It is the modern version of Big Brother.

I don’t like that. Some people thought Facebook was worth almost US$100 billion. I don’t think it is and neither, apparently, does “the market”.

The algorithm required to build a web page is trivial compared to alien-hunting. The storage capacity per user is tiny compared to a typical user’s home storage, and as the need to push for extra advertising revenue increases it is inevitable that people will become alienated and want out.

Finally, unlike a bank, Facebook is about trivia, and if it “goes down\” the world doesn’t end. So it is a very safe application to work on. Indeed if everyone who used Facebook left their computers on 100% of the time, software to emulate Facebook’s core activities would be fairly simple to implement.

For each friend, a local application could bug all your friends\’ computers for recent “activity”, and build up the relevant web page for display without any market bias or promotional activities. Bliss!

The problem is that most of us turn our computers off. So that complicates any would-be programming model. An alternative is therefore for users to lease space for their private data on “the cloud”, and allow friends with an appropriate key to see what you have been up to.

Another option might be to have a micro PC like the amazing 5 Watt Raspberry Pi permanently online in your house that contains all of your social media data.

The Raspberry Pi costs A$35, and only costs about 2c/day to run.

As users become increasingly uncomfortable with Big Brother knowing things about them that they’d prefer were kept secret, programmers will find ways to harness the mother of all supercomputers – the massively distributed global internet and all the home computers that reside on it.

The community-run network Diaspora* is one attempt to do this for Facebook but other applications will follow.

When someone gets one right they might not discover aliens, but the effect on our lives might be almost as profound.