ARCHER Outreach

ARCHER is a Supercomputer run by EPCC which is based just outside Edinburgh.

The ARCHER Outreach team works to provide information about Supercomputing to schools, colleges, Science Festivals, and anyone who wants to learn about what a Supercomputer is and what it can be used for.

If you hover your mouse over any text underlined like this then additional information and definitions will be displayed.

Supercomputing

What is a "Supercomputer"?

Supercomputers differ from standard computer in that they have 1000s of processors instead of just one or two. Supercomputers can perform calculations that would not be possible on a standard desktop.

 

ARCHER
 
Photograph of ARCHER supercomputer
A Cray XC-30 Supercomputer.

 

Hardware

Processors

ARCHER compute nodes each contain two, 12-core processors running at 2.7 GHz.

 

An ARCHER compute blade
 
Photograph of an ARCHER compute blade
One blade holds four nodes, each node has two processors.
The shiny copper heatsinks help keep the processors cool.

 

There are 118,080 processor cores in total on 4920 nodes.

In theory the whole of ARCHER could perform 3 x 1014 instructions per second.

 

An ARCHER compute blade
 
Photograph of an ARCHER compute blade being slotted into a cabinet
about to be slotted into the cabinet.
Each cabinet can hold up to 48 blades.

 

 

Bits and bytes and things...

A bit is either a 1 or a 0.
A byte is 8 bits - that's enough to store an integer value between zero and 255
A kilobyte, written as KB, is 1024 bytes - that's enough to store a few pages of simple text
A megabyte, written as MB, is 1024 KB or 1,048,576 bytes - that's enough to store a photo from a digital camera
A gigabyte, written as GB, is 1024 MB or 1,073,741,824 bytes - that's enough to store a full length movie
A terabyte, written as TB, is 1024 GB or 1,099,511,627,776 bytes
A petabyte, written as PB, is 1024 TB or 1,125,899,906,842,624 bytes
An exabyte, written as EB, is 1024 PB or 1,152,921,504,606,846,976 bytes

 

Memory

ARCHER has 64 GB of memory per node, giving a total of 307.5 TB of memory.
Unlike a desktop computer where the whole memory is available to the single processor, each 64 GB of memory is only available to the 2 processors of that node. However, specialist programming tools do allow data held in one node's memory to be accessed by other nodes.

Backing Storage

The Research Data Facility (RDF) consists of 7.8 PB disk, with an additional 19.5 PB of backup tape capacity.

Power consumption

At zero load (computer is switched on but not running any jobs) the compute nodes on ARCHER draw approximately 400 kW of power and at full load they draw approximately 1200 kW of power.

The ARCHER cabinets are kept cool by pumping water through in cooling pipes. The water enters at approximately 18°C and, after passing through the cabinets comes out at around 29°C. The water can be cooled back down and re-circulated. When necessary, the water is cooled by electrical chillers but, most of the time, ARCHER can take advantage of the cool Scottish climate and cool the water for free simply by pumping it through external cooling towers, so saving significant amounts of energy.

 

Who uses them and why?

Supercomputers are used for research and simulations.

 

For example:


     Cloudy sky     Turbulent flow     Dinosaur skeleton


You can read about more examples of research on supercomputers here: http://www.hector.ac.uk/casestudies.
HECToR was the predecessor of ARCHER - much of the work started on HECToR is still ongoing on ARCHER.

 

These tasks couldn't be run, even on the best desktop computer available. Some tasks would run, but would take years to complete (not much use for forecasting tomorrow's weather!). Other tasks couldn't be run at all because so much data needs to be held in memory simultaneously e.g. the atmospheric pressure reading for every square metre of the earth's surface.

 

What not to use a supercomputer for.

Supercomputers are not suited to every computing task. User interaction on a supercomputer is very limited. Usually tasks are sent to the supercomputer where they run when processors are available. Playing a computer game on ARCHER would be rubbish. It might take several minutes to receive a response back from ARCHER, whereas when you play a game on your desktop computer the game updates many times per second.