Understanding ‘big data’

Understanding ‘big data’

ROMA – KHOBATHA Setetemela, a Computer Engineering Lecturer at the National University of Lesotho (NUL), has devised and successfully tested a system to help scientists simplify high performance computing to enable processing of Big Data.
In short, Setetemela’s system allows an average scientist to program a complex computer system using a very simple language without having to understand the “nuts and bolts” of that complex system.

His work, in which he collaborates with the University of Cape Town (UCT), is part of one of the biggest and most bewildering multinational scientific projects of all times—the building of a Square Kilometer Array (SKA), a titanic Radio Telescope which will be the first of its kind.
Today’s world is a world of big data.

That is because, “with so many devices now connected to the internet, there has never been a time when so much information about so many things has been collected’’.

SKA will be just one such system where astronomical amounts of information will need to be collected and interpreted in the blink of an eye.
But the application of the system Setetemela is developing applies well beyond the SKA.
As we navigate this strange terrain, we have no choice but to speak and think in parables in order to understand the significance of what Setetemela is doing for the world of science.

Imagine three people, Thabo, Thabiso and Dominique.
Thabo wants to convey a message to Dominque but he can only speak Sesotho and Dominique can only understand French.
Fortunately, there is Thabiso who does not only understand both languages but he is also an interpreter.

He listens to Thabo’s message in Sesotho, processes it and conveys the message to Dominque in French.
That is Setetemela’s solution in a nutshell.

We will soon identify what is the message and who are Thabo, Thabiso and Dominque.
But let’s start here.

“When data gets literally big, big in volume and big in variety, we need high performance computers (the strong ones), to make sense out of that data,” Khobatha says.

These big processing machines rely both on high performance software and high performance hardware (nota bene, if your brain is hardware, your mind is software).

One of the modern high performance hardware is called a Field Programmable Gate Array (FPGA).
It is just one of the hardware and it is an amazing one at that. It is hardware that you can program such that it changes from one form to another to suit your needs much like the strangeness of using your mind to change your brain.

More importantly, it can process many pieces of information at the same time, hence its importance in high performance computing.
“Its capacity to change from one form to another is odd because we are more familiar with the kind of hardware that just won’t change, the so-called General Purpose Processors such as the CPUs in your laptops,” Setetemela says.

In your laptop, you can’t programme I-7 to become I-8.
If you need I-8, you have to buy a new computer.
“FPGA is much more flexible,” he says.
“For instance, you can change an FPGA from being a circuit that multiples numbers to one that encrypts (converts) information.”
No doubt you can already feel that we are now talking of something very powerful here.
But along with that power, comes the complexity.

It is not easy to program FPGA. In fact it is very difficult, Khobatha says, and that is the heart of his research, to make the complex simple.
But this challenge is nothing new. In the early days of CPU, only the best and the brightest used to be able to develop computer programs.
In time, new languages emerged, to make easy for the “rank and file” scientists to program hence the development of the likes of C++, Java, Python, etc.

The biggest challenge with FPGA is that there are no defined standardised languages to program it.
Instead, new tools are developed based mainly on intuition.
That is where Setetemela comes in.

He has developed a logical framework, which is not perfect, of course, through which such tools can be developed and he tested it successfully.
Let’s look at how it worked.

He took an application called Monte Carlo (option pricing, for those who know financial computing).
If you still remember our Thabo, Thabiso and Dominique example, Monte Carlo is the message Thabo wants to convey to Dominque.
But that message is carried in a language called Python (Sesotho) which Dominique (FPGA) does not understand.

So instead of Thabo communicating directly to Dominique, he conveys the message using a tool called Migen (Thabiso’s mind), which does not only understand Sesotho (python, as simple programming language) but can translate Sesotho (python) to French [FPGA code], a language understood by Dominique!

In short, Setetemela’s logical framework allows an average scientist to program an otherwise a very difficult-to-program.
FPGA using a very simple language such as python without having to understand the nuts and bolts of FPGA — in the same way a driver does not need to understand how his Mercedes Benz engine functions before he can start driving.

Own Correspondent

Previous Prisons boss grilled
Next Mohair farmers win interim order

About author

You might also like

News

DC showdown looms

Staff Reporter MASERU – A showdown is looming at the Democratic Congress (DC)’s leadership conference this weekend as intra-party squabbles intensify. Daggers are already drawn between two factions that have

Local News

Tender scandal rocks army

MASERU – THE Lesotho Defence Force (LDF) awarded a huge tender to supply military attire to two unregistered companies two years ago, thepost heard this week. The revelation was disclosed during

Local News

NGO pledges to change lives of the poor

MAPOTENG – A HUMANITARIAN non-governmental organisation has pledged to bring change to the lives of more than half a million boys and girls in the next five years, as it seeks