The Great White South - Linux in Antarctica

This article first appeared in LinuxUser & Developer in 2003

The Antarctic is the last wilderness, remote, hostile, and uninhabited, except by penguins and scientists. The attraction of the Antarctic to scientists is the very wilderness that keeps tourists and colonists at bay. The penguins are there because the South Atlantic is where they belong, down among the fish, the squid and the krill. For the scientists, Antarctica is a key element in the fabric of the world's ecosystem. The Antarctic icecap contains almost 70 per cent of the world's freshwater and ninety per cent of the world's ice. Huge icebergs break off each year from the floating ice shelves and half of the surrounding ocean freezes over in winter, more than doubling the size of the continent.

The processes taking place in the Antarctic affect the world's climate and its oceans, linking the continent inextricably to what we experience thousands of miles away. In understanding global change the Antarctic has a crucial role to play. Locked up in its 4 km thick ice sheet is a record of past climate for the last 500,000 years. Trapped bubbles in the ice hold an archive of atmospheric gases, and evidence for levels of global pollution by industry, volcanic eruptions, agriculture and atomic bombs is frozen into the ice. Equally important is the evidence for ozone depletion in the upper atmosphere - The discovery by the British Antarctic Survey (BAS) of the Antarctic ozone hole (caused by emissions, mainly in the northern hemisphere, of chlorofluorocarbons (CFCs) and halons, used in refrigeration and industrial solvents), provided an early warning of the dangerous thinning of the ozone layer worldwide, and spurred international efforts to curb the production of CFCs.

The British Antarctic Survey is based in Cambridge, employs over 400 staff, and supports three research stations in the Antarctic, at Rothera, Halley and Signy, and two stations on South Georgia, at King Edward Point and Bird Island. The Antarctic operations are sustained by two ice-strengthened ships. RSS James Clark Ross has advanced facilities for oceanographic research. RSS Ernest Shackleton is a logistics ship used to bring supplies to the stations. Four Twin Otter aircraft fitted with wheels and skis are operated from Rothera and Halley, and a Dash-7 aircraft provides transport between Rothera and the Falkland Islands. Most research establishments specialise in one or two scientific disciplines. Because Antarctica is such an important source of scientific information the British Antarctic Survey covers everything from the biological sciences to atmospheric physics.


Penguin Island

For the early explorers Antarctica was a forbidding environment. Less than a century ago the technology available to the likes of Amundsen, Scott and Shackleton was minimal. Shackleton disappeared south in 1914. His ship was crushed by the ice floes of the Weddell Sea, and his men huddled under upturned boats on the coast of Elephant Island while he undertook one of the greatest small boat journeys ever, climbed the unclimbable mountains of South Georgia, and reached the whaling stations at Husvick against all odds - achieving in the process the ultimate rescue of his companions, whose plight was unknown to the outside world. Fortunately for the scientists and support staff of the British Antarctic Survey who live for many months of the year in small groups of huts on the small outcrops of Rothera, Halley and Signy on the outer reaches of the Antarctic continent, or at Bird Island and King Edward Point on South Georgia, the outside world is now little more than a ship's journey, a plane flight, or an email away.

The scientists huddled on the rocks or ice shelfs of South Georgia or the Antarctic rim now inhabit tidy cabins and laboratories, equiped with laptops, electric lights and the other gizmos of modern life. The BAS personnel at Rothera or Halley or Bird Island and the other outposts of the British Antarctic Survey use an email messaging system to maintain contact with the outside world that was developed in-house, with customised data compression, and runs on a GNU/Linux infrastructure. The BAS Antarctic Message System is a store-and-forward email system that transmits over an Inmasat satellite link between the ships and bases and the main base in Cambridge. Andy Wood, senior systems consultant for the BAS, says: "We don't have 24 hour connections to our bases and ships as yet, so we store up email in Cambridge and blast that out across high speed dialogue lines 3 or 4 times a day when the various bases want to exchange email." There is a delay in receipt and sending of mail due to the limitations and cost of satellite time, but BAS are installing a new satellite link that will give the ships and bases 24/7 connection, data transmission capabilities, and voice-over-ip. There is at least one Linux server at each of the bases and on each of the ships of the British Antarctic Survey.

An earlier version of the messaging system ran on Digital's Vax/VMS "but at Bird Island we had a Debian Linux system installed in 1997, on a laptop connected to a dialogue line over Inmasat", Wood recounts. Bird Island lies off the north-west tip of South Georgia, and is home to vast number of birds, including 50,000 breeding pairs of penguins. Six to eight people usually work at the station during the summer, and four remain for the winter, undertaking research programmes into seabird and seal population dynamics, feeding ecology and reproductive performance. "Although we had Vax/VMS at most of the bases, for Bird Island it was overkill for the small amount of people there. We were looking for a small laptop solution. Linux was seen as a possibility - we looked at RedHat and Debian, and finally came down on Debian because of the superior administration tools it provided at that time - dselect gave us all the dependencies we needed for installing applications. Our current installation at Bird Island is a Linux box that we put there in 1999, that runs a Samba file server, an Apache web server, and an early version of Oracle8 for Linux." And Debian GNU/Linux churns away through the long nights of the South Atlantic winter...

Changing the landscape

The success of the initial installation of Debian at Bird Island encouraged further deployments of Linux, and Linux now operates in many different guises throughout the IT infrastructure of BAS, in its headquarters in Cambridge, on its two ships, RSS James Clark Ross and RSS Ernest Shackleton, and at its five scientific survey bases that are widely dispersed around the South Atlantic.

For historical reasons, BAS deploys a selection of Netware, Linux, Windows and Unix systems, each of which were chosen for their aptitude for the job that was expected of them. Linux is still relatively new in this landscape, but is beginning to make its presence felt. Jeremy Armitage, the IT manager of BAS in Cambridge, elaborates, "Linux hasn't been around that long, but whenever we are implementing new systems, or updating the systems we already have, we are looking at Linux to see whether it provides us with better functionality or cheaper facilities. Open source software has given us a lot of opportunities to look at very different solutions for the problems we have. The fact that a company such as Oracle is providing software on Linux is a nice incentive for us to go that way."

Linux is still seen primarily as a server side operating system, and is gradually encroaching into the space previously occupied by Solaris. BAS doesn't purchase its Linux machines pre-installed or preconfigured, but relies on homegrown experience to install and configure the system, primarily on HP/Compaq Proliant hardware. The It support system, based in Cambridge, consists of 10 people, 5 of whom support the Unix and Linux systems. Three of the Unix/Linux support personnel run on Linux desktops by preference.

Red Hat has usurped Debian as the distribution of choice, at least on the server side. "When we are evaluating a new service, where it's appropriate, where Linux provides the best solution, we install Linux. The main consideration is the hardware we can run it on."

"The fact that you can run Linux on cheaper hardware is an incentive for us to look at it", according to Wood. Funding for BAS tends to go in cycles. "The last couple of years we have been quite well funded and have been able to install a large storage network, and upgrade our local area network."

The British Antarctic Survey receives its funding from the Natural Environment Research Council (NERC). According to the BAS website the entire science budget for BAS for 2000-2001 was £29.8 million. Of this, the core strategic science programme was budgeted at £6.5 million, and supporting the science, the infrastructure was budgeted at £21.0 million. The latter figure included £18.0 million expenditure on ships, aircraft and research stations. As with any establishment, the IT budget has to be restrained to ensure funds are directed towards the core business of the organisation, which in the case of BAS, is scientific research.

Linux is tending to displace Unix, rather than Windows, partly because historically Unix has been the platform for servers, and Linux on Intel is cheaper. "Linux is becoming the preferred system on the server side, and for workstations" says Jeremy Robst, support engineer for BAS, "but not on the desktop, although in two or three years things might be different".

Much of the internal infrastructure software at BAS is written on and for Linux, including the internal helpdesk software that was written by Robst. Aside from the obvious cost advantages of deploying Linux on the server side, Robst believes that the administration, management, and flexibility are the greatest advantages that Linux offers, at least to the systems administrator. Even where Solaris remains in situ, because "it just works", it is often running Samba or Apache or other Open Source applications. The excellent and informative website of the British Antarctic Survey currently runs on Solaris, although interim web site development is performed on a Linux machine. Plans are afoot to move the website to Linux in the near future.

Beowulf

As with most scientific and research establishments in the modern age, computer technology is essential to the work of the British Antarctic Survey. Fifty years ago scientists were restricted to the use of ink and graph paper to assess and compile their data. The first computer revolution of the seventies and eighties vastly enriched the possibilities. But the second revolution in computer technology that has taken place over the last decade has transformed the work of most scientific establishments. The Internet has made data more available. Data capture has become more refined. Computers can cope with volumes of data that were unimaginable a few short years ago, and can perform iterative calculations that were beyond the realms of possibility even more recently. Moreover, the price/performance ratios that were so forbidding in the past have been transformed by the effects of Moore's Law. For most of us this means little more than the notion that office applications are a little less clunky than they used to be. In the world of science it means that the margins of knowledge can be explored a lot further.

The special contribution of Linux and open source has been the cluster, where any number of instances of Linux running on commodity hardware boxes are able to simulate the performance of a supercomputer. The climate research group of BAS has two Beowoulf clusters which run the Hadley Centre Climate Prediction Model from the Meteorological Office. The implementation has been driven by the scientists themselves. The IT Department put the hardware together, but the scientists themselves built the clusters. BAS has a 13-node cluster (8 dual-board Athlon MP 1600's, with 5 dual-board 2200's recently added, and an older 4-node cluster consisting of dual PIII's 1GHz from Compusys. Armitage predicts that this form of computing will be a growth area within the organisation during the next few years, as scientists find other permutations for the technology.

The Beowulf project was founded by Donald Becker and Thomas Sterling in 1993 when they began "sketching the outline of a commodity-based cluster system designed as a cost-effective alternative to large supercomputers." The first Beowulf cluster was built by NASA and CESDIS scientists based at the Goddard Space Flight Center in Greenbelt, Maryland, to address problems associated with the NASA Earth and Space Sciences (ESS) project. One of the goals of the ESS project was to determine "the applicability of massively parallel computers to the problems faced by the Earth and space sciences community." The Beowulf project is driven by its users, and was the first project of its type, but is being increasingly challenged by COTS implementations of the technology, not all of which are truly open source.

The importance of the BAS climate prediction model can be surmised from a position paper on the BAS website, signed by Dr John King and members of the Physical and Geological Sciences Divisions of the British Antarctic Survey, September 2002. The paper begins:

"Is a changing Antarctic climate signposting more ominous changes to come elsewhere? Like the ozone hole is it giving humanity early warning that more precipitous changes are to come? Or do the observations of ice retreat and warming temperatures lie within expected bounds in a highly variable climatic regime? These are the questions that spur on the meteorologists, glaciologists, geologists and atmospheric physicists who study the Antarctic environment." The conclusions are admirably objective, but the evidence of global warming and disintegration of the Antarctic ice-shelf that is the reault of cumulative research over the last fifty years is decidedly uncomfortable.

On the ship

Conditions in the Antarctic are harsh. Every year since he joined BAS in 1997, Robst has spent two or three months on the research vessel, RSS James Clark Ross, (after the reknowned nineteenth century polar explorer), providing IT support to the scientists. "You read about Shackleton and all the hardships,", he says, "But I just get on a plane and three or four days later I'm sailing around the Antarctic.

As far as is possible the IT centre on RSS James Clark Ross duplicates the facilities back in Cambridge. His first job after boarding the vessel is to ensure that all the machines are tied down in case of storms. But the sea and the weather has remarkable little impact on the machinery. The Antarctic cold is more damaging, and some of the machines in use at the bases are "ruggedised" for protection.

The basic network on RSS James Clark Ross is built around Novell Netware. Linux workstations and servers are used for data logging. "We have a clear split between the scientists in the organisation, and the people who run the organisation, the finance department, procurement etc.", says Wood, "For them the operating system, probably as it should be, is irrelevant. They just need the applications. Unless there's a big push to get all those applications out there on Linux, it's going to be a little while - the market is entrenched. It's not about what linux can do or can't do."

The scientists are much more likely to choose to use Linux on their workstations, and run OpenOffice rather than Word and Excel. The choice is driven by the users themselves. One reason for this is the growing number of open source or public domain applications exchanged by scientists in many different disciplines, which are often writtem on the Linux platform. Another reason is that Linux is sufficiently like Unix to provide a comfortable zone in which to work for those who have grown up with one flavour of Unix or another. The final reason is that some like Linux becuase they like Linux.

"Personally I use OpenOffice", says Robst. "But we used to run WordPerfect before we moved over to Office 2000. The site license for WordPerfect (as an academic institution) was £50, while Office costs thousands. The problem wasn't the functionality, but the .doc format. That was the reason we moved, purely the format. It's crazy to have to change your office suite just for the file format but that's the way things are. One of the reasons I like Free Software is that you don't have to worry about the licenses and license management. I like the whole philosophy of Free Software"

Rothera is the largest of the BAS research stations, and can accommodate up to 128 people during the summer months, falling to 20 during the Antarctic winter, the majority of whom are support staff. The main areas of research are glaciology, geology, geophysics, biology and meteorology. Rothera has a good harbour and an airstrip.

Halley is on the Brunt ice shelf, and operates throughout the year with a maximum population of 65 in the summer and an average of 15 during the winter. In the winter there is darkness for 105 days. Halley is responsible for much of the atmospheric research undertaken by BAS. Supplies are landed twice a year by ship onto the ice shelf and then towed on sledges for several miles from the ice edge.

Signy is the oldest British research station in Antarctica and is built on the site of an abandoned Norwegian whaling station. The base was occupied continuously from 18 March 1947 until 13 April 1996, and has since been occupied only during the Antarctic summer. Research at Signy includes lake chemistry, micrometeorology, microbial community dynamics, catchment-lake interactions and the CCAMLR penguin survey.

Each of these Antarctic bases, and the Bird Island and King Edward Point bases on South Georgia, is running at least one Linux server for most of the year.

A penguin colony

As already indicated, Linux use in the infrastructure and on the workstations of the British Antarctic Survey is on an upward curve. As longtime Novell Netware users, the BAS software support staff are keenly anticipating the opportunity to switch from Novell networking on NetWare to Novell on Linux, which will reduce the number of servers as Linux doubles up on other duties. Equally, there is a mid-to-long term interest in the development of the Linux desktop from the likes of Novell/SUSE and Sun Microsystems.

Open source software is favoured in the scientific community because of the flexibility and opportunity it affords for mutual exchange of knowledge, information and techniques. For a scientist honed in bare knuckle C or Fortran programming, Linux is an easier and more affordable platform to work with than Windows or Unix.

In the immediate future BAS is setting up setting up a research group on bioinfomatics and genomics. Much of the software developed in those fields is already distributed under Free Software Licences, and has been designed to run on Linux with the GNU tools.


First to the pole?

As early as 1993, Andrew Tridgell, lead developer of the Samba team, was approached by Martin Hendy of AUSLIG (Australian Surveying and Land Information Group), to investigate whether Linux would be the appropriate operating system for a project he was undertaking with the Global Positioning System (GPS).
The ensuing deployment led to the placement of Linux boxes at various points around Australia, and at the Mawson, Davis and Casey bases in Antarctica. The systems used the 0.99 version of the Linux kernel and were honed down to fit in their entirety on a 1.44 megabyte floppy disk. The Linux machines were attached via serial ports to a satellite receiver system, which monitored the 32-satellite GPS, and downloaded the relevant data to the Linux systems, before transmission to a base in Canberra.
The Linux systems in Antarctica ran Digital 48633DX MT PCs with 8MB RAM and two 345MB IDE hard disks, one in use with the 'production' system, and the other identically configured as a backup.
One reason for the adoption of Linux for this system was that it was possible to completely replace the system remotely, via modem or the net. The Linux root partitions were 20MB and "only half full". Tridgell explained, "Most of them are, in fact, taken up with things that are only included just in case, like kernel sources, gcc, sources for the TurboRogue (satellite) controlling software, and Emacs 19.16 (to make life a little easier)."
The first systems in Antarctica came on-line in October 1993.

Richard Hillesley




Back to top