Throughout history, technology has gradually introduced its own terminology. Some of the words we use in IT today are often unfamiliar to the general public, but are needed to describe the mind-boggling power of the computing technology of Google, Microsoft, Amazon or IBM.

Let’s talk about size

We need increasing units of measurement to address the ‘hyperscale’ of the sheer numbers involved in super-computing. We move through steps which increase a thousand-fold. Most people will recognise terms like Kilo, Mega, Giga, & Tera, but this is probably where the average personal computer user currently tops out. There are Terabyte drives available for less than £25. That equates to a million Megabytes – or 1,000,000,000,000 bytes.

To get closer to the computing ability of big data operators, we have to climb further through Peta, Exa and Zetta, taking us up in steps of 1,000 each. The next level up is Yotta, landing at a million million million million bytes. As far as we know, the world has not yet seen a Yottabyte of data. It is currently estimated that the entire data of the world amounts to about 1.2 Zettabytes.

Mind-boggling operating ability

We use the same thousand-fold prefixes for machine operations, which are referred to as “FLOPS”. This means Floating Point Operations Per Second. This is a description of the processor power in realistic terms rather than simple instruction cycles.

Although the costs of funding these massive data centres is obviously climbing as the entire infrastructure expands, the comparative costs per Gigaflop has diminished exponentially from 1.1 trillion dollars in 1961 to around 8 cents in 2015.

Examples of computing power

These days, big data corporations are usually talking at Peta level to describe their  accomplishments – that’s 1,000,000,000,000,000,000,000 bytes of data.

  • China has a 93 Petaflop computer, called the Sunway TaihuLight.
  • In 2012, Facebook’s largest Hadoop cluster had about 100 Petabytes.
  • In 2013, Microsoft moved their Hotmail accounts to Outlook – transferring about 150 Petabytes of data in 6 weeks.
  • One Petabyte of MP3 music would take about 2,000 years to play.
  • In comparison, some scientists believe the human brain might be able to hold something like 2.5 Petabytes of data.

Systems growing smarter

In 2013, Google and NASA joined in collaboration to produce something called the D-Wave X2 Quantum Computer, which they claim is 100 million times faster than conventional computer chips. Certain algorithms can now be processed in seconds when these used to take years.

Google doesn’t say how much computer power it has, but the best estimates hover around 10-15 Exabytes of data. This would make Google about as smart as the cumulative brain power of the entire population of the principality of Liechtenstein, one of the smallest countries in Europe.

Examples of infrastructure

In some ways, it would make more sense to talk not in terms of bytes or flops, but buildings.

  • Amazon is estimated to have around two million servers. It has eleven cloud regions which each have multiple data centres totalling 28 sets globally, some of which contain 50 – 80 thousand servers.
  • Google has three regions with eight sets. They spent $2.4 billion on data centres in 2007, but by 2015 this had gone up to $11 billion and in the first quarter of 2016 alone they spent $2 billion more.
  • Microsoft has 17 regions and claims over a million servers.
  • Facebook doesn’t say how many servers it has, but was initially hosted on one single server and then expanded. It grew initially by hiring data centres, but is now building its own. The size of such a campus can be around 300,000 square feet.

Electricity requirements

The figures for power consumption are overwhelming and this sort of power involves a carbon footprint that is also considerable. The New York Times estimated that in 2010 Google would have used about one quarter of the power of an entire nuclear power plant.

(Google’s data centres use around 260 megawatts of power, or, roughly around 0.01 of the world’s energy.)

Google has pledged to be a carbon-neutral company, but before it offset its emissions in 2010 it generated 1.46 million tons of CO2. We may not consider the cost of our Google searches as it is not on our electricity bill, but Google says that 100 searches would be equivalent to running a 60 Watt bulb for 28 minutes. They do however claim that choosing Gmail instead of a locally hosted server actually would mean an annual individual power saving of 170 kWh, thus reducing the average carbon footprint by up to 100kg of CO2.

And in case you are wondering what’s going on commercially are behind the scenes, Google employs about 57,000 people, and the company is worth around $498 billion. Facebook and Amazon have a market value of about $340 billion.